Volume 14. No

Transcription

1 Volume 4. No 03

2 Transporta un sakaru institūts (Transport and Telecommunication Institute) Transport and Telecommunication Volume 4, No - 03 ISSN ISSN (On-line: Riga 03

3 05 T8 EDITORIAL BOARD: Prof. Igor Kabaskin (Editor-in-Cief), Transport & Telecommunication Institute, Latvia Prof. Irina Yatskiv (Deputy/ Managing Editor), Transport & Telecommunication Institute, Latvia Prof. Adolfas Baublys, Vilnius Gediminas Tecnical University, Lituania Dr. Brent D. Bowen, Purdue University, USA Prof. Ernst Frankel, Massacusetts Institute of Tecnology, USA Prof. Arnold Kiv, Ben-Gurion University of te Negev, Israel Prof. Eugene Kopytov, Transport & Telecommunication Institute, Latvia Prof. Aleksandr A. Kondratiev, Nort-West State Tecnical University, Russia Prof. Andrzej Niewczas, Lublin University of Tecnology, Poland Prof. Lauri Ojala, Turku Scool of Economics, Finland Prof. Juri Tolujew, Transport & Telecommunication Institute, Latvia Prof. Wolfgang Kersten, Hamburg University of Tecnology, Germany Prof. Cristina Pronello, Polytecnic Torino, Italy Dr. Vadim Doncenko, Scientific and Researc Institute of Motor Transport, Russia Dr Andrius Jarzemskis, Vilnius Gediminas Tecnical University, Lituania Dr. Olegas Prentkovskis, Vilnius Gediminas Tecnical University, Lituania Prof. Roberto San José García, Tecnical University of Madrid, Spain Host Organizations: Transport and Telecommunication Institute, Latvia Boriss Misnevs, Rector Telematics and Logistics Institute, Latvia Igor Kabaskin, Director Supporting Organizations: Latvian Transport Development and Education Association Latvian Academy of Sciences Latvian Operations Researc Society Telecommunication Association of Latvia Latvian Academic Park Pre Press Manager Elena Rutkovska, Transport & Telecommunication Institute, Latvia Language Editors Lucija Paegle, Transport & Telecommunication Institute, Latvia Tecnical Editors Olga Davidova, Transport & Telecommunication Institute, Latvia All articles are reviewed. Articles can be presented in te journal in Englis. EDITORIAL CORRESPONDENCE Transporta un sakaru institūts (Transport and Telecommunication Institute) Lomonosova iela, LV-09, Riga, Latvia. Pone: (+37) Fax: (+37) ttp:// TRANSPORT and TELECOMMUNICATION, 03, Vol. 4, No ISSN Te journal of Transport and Telecommunication Institute (Riga, Latvia). Te journal is being publised since 000. Te papers publised in Journal Transport and Telecommunication are included in SCOPUS (since 008, Vol. 9, No ), Elsevier Database, Te Summon, VINITI, Versita, Transportation Researc Board, JournalTOCs, INSPEC, TEMA (Tecnology and Management) ttp:// Copyrigt Transport and Telecommunication Institute, 03 ii

4 Transport and Telecommunication Vol. 4, No, 03 CONTENTS Exploratory Assessment of te Limiting Extended Kalman Filter Properties Constantinos Antoniou, Alexandra Kondyli, Georgia-Maria Lykogianni, and Elias Gyftodimos... Telematics Business and Management in Barain Market Marwan Raje Hussain, Abduljalil Zainal, Wael Moamed Elmedany, Moamed Waleed Fakr... 3 Planning and Financing Scemes Linked to te Decision-Making for te Interconnection of Long-Sort Distance Transport Eftiia Natanail, Giannis Adamos... 0 Distinguising between Spatial Heterogeneity and Inefficiency: Spatial Stocastic Frontier Analysis of European Airports Dmitry Pavlyuk... 9 Mobile Autonomous Robots a New Type of City Public Transport Vasili Sut, Valery Kasyanik Individual Attitudes and Sopping Mode Caracteristics Affecting te Use of E-Sopping and Related Travel Federica Crocco, Laura Eboli, Gabriella Mazzulla Some Approaces to te Rolling Weels Dynamics Modelling in te Weigt-In-Motion Problem Alexander Grakovski, Yuri Krasnitski, Igor Kabaskin, Yuri Sikerzitski, Victor Truacov Autors index Personalia Cumulative Index Preparation of Publications... 9 iii

5 Transport and Telecommunication Vol. 4, No, 03 Transport and Telecommunication, 03, Volume 4, No, Transport and Telecommunication Institute, Lomonosova, Riga, LV-09, Latvia DOI 0.478/ttj EXPLORATORY ASSESSMENT OF THE LIMITING EXTENDED KALMAN FILTER PROPERTIES Constantinos Antoniou, Alexandra Kondyli, Georgia-Maria Lykogianni and Elias Gyftodimos National Tecnical University of Atens, Greece, Scool of Rural and Surveying Engineering Laboratory of Transportation Engineering Zografou, GR-5780, Greece P.: Fax: antoniou@central.ntua.gr, akondyli@gmail.com, gmlikog@gmail.com elias.gyftodimos@gmail.com Most of te metodologies for te solution of state-space models are based on te Kalman Filter algoritm (Kalman, 960), developed for te solution of linear, dynamic state-space models. Te most straigtforward extension to nonlinear systems is te Extended Kalman Filter (EKF). Te Limiting EKF (LimEKF) is a new algoritm tat obviates te need to compute te Kalman gain matrix on-line, as it can be calculated off-line from pre-computed gain matrices. In tis researc, several different strategies for te construction of te gain matrices are presented: e.g. average of previously computed matrices per interval per demand level and average of previously computed matrices per interval independent of demand level. Two case studies are presented to investigate te performance of te LimEKF under te different assumptions. In te first case study, a detailed experimental design was developed and a large number of simulation runs was performed in a syntetic network. Te results suggest tat indeed te LimEKF algoritm is robust and wile not requiring te explicit computation of te Kalman gain matrix, and tus aving vastly superior computational properties its accuracy is close to tat of te exact EKF. In te second case study, a smaller number of scenarios is evaluated using a real-world, large-scale network in Stockolm, Sweden, wit similarly encouraging results. Taking te average of various pre-computed Kalman Gain matrices possibly reduces te noise tat creeps into te computation of te individual Kalman gain matrices, and tis may be one of te key reasons for te good performance of te LimEKF (i.e. increased robustness). Keywords: Origin-destination flows estimation and prediction, state-space models, Extended Kalman Filter (EKF), Limiting Extended Kalman Filter (LimEKF) Introduction Effective management of large-scale transportation systems requires te support of detailed traffic simulation models. Suc traffic simulation models involve a large number of parameters and input data tat must be calibrated to matc surveillance data. Depending on te scope, calibration can be eiter performed off-line or on-line. Te role of off-line calibration is to develop a istorical database tat supports te simulation model s ability to replicate average or expected traffic conditions as captured by arcived surveillance data. Te output of tis process is expected to perform adequately in off-line, planning-level studies, suc as te evaluation of alternative traffic management strategies. Traffic conditions are affected by a number of exogenous factors, including incidents and weater conditions. In order to capture tese fluctuations, terefore, te off-line, average, parameters must terefore be updated dynamically []. Te real-time performance of te traffic simulation models could terefore be significantly enanced by re-calibrating te a priori parameter estimates using te available surveillance data. Off-line and on-line calibration as similar objectives, and a general formulation tat encompasses bot as been proposed by Antoniou et al. []. Off-line calibration is a global optimisation step aiming to determine te mean values of te parameters and inputs, wile on-line calibration adjusts te off-line values to reflect prevailing conditions. Data from prior days may be available for off-line calibration, wile on-line calibration will generally use smaller amounts of recent, real-time data. Furter, off-line calibration as no explicit computational constraints, wile on-line calibration as to be faster tan realtime. Finally, off-line calibration must estimate all model inputs and parameters, wile on-line calibration could focus on a subset of critical parameters. Te remainder of tis paper focuses on on-line calibration.

6 Transport and Telecommunication Vol. 4, No, 03 Several approaces for on-line calibration of dynamic traffic simulation models ave been proposed. Tavana and Mamassani [3] use transfer functions (bivariate time series models) to estimate dynamic speed-density relations from detector data, wile Huyn et al. [4] extend tis work by using transfer functions in a simulation-based DTA framework. Qin and Mamassani [5] evaluate te same model wit sensor data from several links in Irvine, CA. Antoniou et al. [6] formulate te problem of on-line calibration of te speed-density relationsip as a flexible state-space model and present solution approaces, validating and comparing tree of te presented solutions [Extended Kalman Filter (EKF), Iterated EKF, and Unscented Kalman Filter (UKF)]. Wang and Papageorgiou [7] and Wang et al. [8, 9] present a general approac for real-time freeway traffic state estimation, wit detailed case studies. Asok and Ben-Akiva [0, ] formulate te real-time OD estimation and prediction problem as a state-space model and solve it using Kalman Filters. Te approac as been implemented in DynaMIT [ 4]. Bierlaire and Crittin [5] present an efficient solution algoritm for OD estimation, wile Zou and Mamassani [6] develop similar Kalman-filter based adaptive OD estimation and prediction procedures using a polynomial trend filter to recursively capture demand deviations from a priori demand estimates. Antoniou [7] develops an approac tat jointly formulates te on-line calibration problem as a statespace model comprising transition and measurement equations. Applicable solution algoritms are presented and compared in Antoniou et al. [8]. Te remainder of tis paper is structured as follows. Te next section provides some background to te problem, introducing te Origin-Destination estimation problem, wile te following section outlines te metodology, leading to and presenting te optimisation algoritms tat are used to solve te state-space formulation of te OD estimation and prediction problem. Two case studies are presented in te following section. Te first one includes several experiments undertaken in a syntetic network, followed by a case study using a real-world large-scale network. Te concluding section provides an overview and discussion of te results obtained, as well as future researc directions. Background Te proposed metodology builds upon te OD estimation and prediction framework presented by Asok and Ben-Akiva [0, 9] and formulated as a state-space model. A state-space model is formulated as a set of: Transition equations; and Measurement equations. Denoting by X te vector representing te number of veicles between eac OD pair departing teir origins during time interval, te transition equation can be expressed in matrix form as: p X+ = f X p + p= q ˆ w, () were ˆX + is an estimate of X +, f p is te matrix of effects of X p on ˆX+, w is a vector of Gaussian, zero-mean, uncorrelated errors, and q is te degree of te autoregressive process. Te measurement equation, wic relates unknown OD flows to observed link counts, can tus be stated in matrix form as follows: Y = A X + v, () were A = a is te assignment matrix mapping te drivers tat departed in interval to te link counts observed in interval and v is te vector of measurement errors, assumed to be zero-mean, uncorrelated as well as uncorrelated wit te transition errors w. Metodology Most of te metodologies for te solution of state-space models are based on te Kalman Filter algoritm [0], developed for te solution of linear, dynamic state-space models. Te most straigtforward extension to nonlinear systems is te Extended Kalman Filter (EKF, sown in Algoritm ), wic employs a linearization of te non-linear relationsip []. Tis algoritm can be applied to simulationbased systems (wic do not ave an analytical representation) by performing te linearization step using numerical derivatives. Te use of numerical tecniques requires a large number of function evaluations (eac implying a run of te simulator) tus making tis tecnique potentially computationally expensive.

7 Transport and Telecommunication Vol. 4, No, 03 3 Initialisation P P X X = = For = to N do Time update T Q F P F P X F X + = = Linearization * * *) ( = = X x x x H Measurement update ( ) ( ) [ ] = + = + = T T P G H P P X Y G X X R H P H H P G End for Algoritm. Extended Kalman Filter (EKF) In Antoniou [7] and Antoniou et al. [8] te Limiting EKF (LimEKF, sown in Algoritm ), a new algoritm tat obviates te need to compute te Kalman gain matrix on-line was proposed. Generation of limiting Kalman gain matrix G and H Initialisation P P X X = = For = to N do Time update T Q F P F P X F X + = = Measurement update ( ) [ ] = + = GHP P P X G Y X X End for Algoritm. Limiting Extended Kalman Filter (EKF) Te performance of tat approximate algoritm proved very close to te exact EKF and raised interest for te approac. In tat work, te off-line computed Kalman gain matrix was te average of all available pre-computed suc matrices. In tis researc, several different strategies for te construction of te gain matrices are presented: e.g. average of previously computed matrices per interval per demand level and average of previously computed matrices per interval independent of demand level.

8 Transport and Telecommunication Vol. 4, No, 03 Platform Implementation Te experimental framework was implemented as a standalone Java application, consisting of tree main components: a) An implementation of te parameter estimation algoritm; b) a wrapper of te traffic simulation; c) te main module wic initialises te experimental set-up and orcestrates te execution of simulation and estimation for te appropriate iterations. Te parameter estimation currently includes implementations of te EKF and Limiting EKF algoritms, and is designed to be modular and easily extensible. Te input to te EKF algoritm includes te initial values of te X vector, wic correspond to te demand values in te traffic network. Also given are te matrices F, P, Q and R, and te vector Y corresponding to te network flow measurements. Te output of te algoritm implementation is mainly te estimated vector X* and te simulated Y* = f(x*). Te implementation of te Limiting EKF is similar, and takes te additional input of te pre-calculated H matrix. Te simulation wrapper implementation abstracts te execution of te traffic simulation software as a function Y = f(x). Te orcestration module implements te main metod, wic reads te configuration of te experiment (specifying wic estimation algoritm is used, providing te estimation parameters, defining te parameters of te mezzo simulation). Te estimation algoritm is executed for te specified time intervals, and te estimation output is persisted to te file system for furter processing. Mezzo [, 3] is mesoscopic traffic simulation software developed by te Centre for Traffic Researc of te Stockolm Royal Institute of Tecnology (KTH). It allows te definition of a road network over wic aggregated traffic is simulated on te level of network links, according to specified values of traffic demand. Te software facilitates te execution of simulations in sequential time slices wit te possibility of varying te demand for eac of te time intervals. Mezzo supports modelling of traffic features suc as traffic signals and road incidents. Te software can be executed eiter wit a grapical user interface allowing user interaction and step-by-step monitoring of te simulation process, or in batc mode as a standalone application. Te simulation results are logged in appropriately formatted text files containing relevant information suc as te (average) link travel times, origin/destination outputs, speeds and densities. Mezzo is developed in C++ and is free and open source under Gnu Public License v. 3. Case Study : Syntetic Network A detailed experimental design was developed and a large number of simulation runs was performed. In order to be able to explore te specific properties of te Gain matrices in a practical way, a small syntetic network is considered, wit varying levels of demand and number of OD pairs. A simulation framework is developed, tat implements te various algoritms and uses te Mezzo mesoscopic traffic simulator. Syntetic Network Description For te purposes of tis researc, a simple syntetic network was developed to explore te specific properties of te EKF and Limiting EKF algoritms. Te syntetic network wic was modeled in Mezzo simulator is comprised of ten links and tree OD pairs (one origin and tree destinations). Te simulation would run for two ours (eigt consecutive 5-min intervals), were te first our was considered to be te warm-up period. To avoid congestion and spillbacks into te network links, te input demand was set at medium levels. Te demand during te warm-up period remained constant, and it varied sligtly at eac subsequent interval. In tis syntetic network it was assumed tat detector data (flows) were available on two of its links. Experimental Design Several scenarios were developed in tis researc to evaluate te EKF and te different Limiting EKF variants using te syntetic network described earlier. More specifically, one of te parameters tat varied trougout te simulations was te OD demand. Five levels of te demand were evaluated; correct knowledge of te OD demand, and its variation by ±0% and ±0%. Additionally, to account for te variability in real life data tat is typically induced due to detector errors or misreading, we furter assumed tat te traffic counts on te two links would vary by 0%, 0%, and -0%. Table summarizes te different scenarios tat were evaluated troug te use of te syntetic network. 4

9 Transport and Telecommunication Vol. 4, No, 03 Table. Different scenarios and parameter levels for te EKF and Limiting EKL algoritm evaluation OD demand variation Traffic counts variation Scenario 0% 0% Scenario 0% +0% Scenario 3 0% -0% Scenario 4 +0% 0% Scenario 5 +0% +0% Scenario 6 +0% -0% Scenario 7-0% 0% Scenario 8-0% +0% Scenario 9-0% -0% Scenario 0 +0% 0% Scenario +0% +0% Scenario +0% -0% Scenario 3-0% 0% Scenario 4-0% +0% Scenario 5-0% -0% Imperative in te estimation of te Kalman Gain matrix (G ) is te measurement matrix H, wic is obtained troug an intermediate linearization step (see Algoritm ). For te EKF algoritm, a % of te linear delta factor was assumed, indicating te percentage of te demand to be considered during te linearization step. To overcome te computationally intensive step of te linearization in te EKF algoritm, te Kalman gain matrix G is replaced wit its limit G (Cui and Cen, 999), for te Limiting EKF algoritm. Tus, te main component of te gain matrix (i.e. te time-dependent matrix H ) is replaced erein wit an average matrix H of a number of available matrices. We distinguis tree cases for computing te average matrix H: (a) it is te average matrix for eac time interval for te scenarios wit varying OD demand ±0%, (b) it is te average matrix for eac time interval for te scenarios wit varying OD demand ±0%, and (c) it is te average matrix across all time intervals for te scenarios wit varying OD demand ±0%. Furtermore, for te implementation of te Limiting EKF it was furter assumed tat te covariance matrices Q and R are identity matrices. Results Te results suggest tat indeed te Limiting EKF algoritm is robust and wile not requiring te explicit computation of te Kalman gain matrix, and tus aving vastly superior computational properties its accuracy is close to tat of te exact EKF. Taking te average of various pre-computed Kalman Gain matrices possibly reduces te noise tat creeps into te computation of te individual Kalman gain matrices, and tis may be one of te key reasons for te good performance of te Limiting EKF (i.e. increased robustness). Tis section demonstrates te ability of te two algoritms to estimate traffic flow conditions in a syntetic network. To evaluate te fit of te estimated and te simulated counts, two measures of effectiveness were considered, namely te normalized root mean square error (RMSN) and te mean percentage error (MPE). RMSN N n= = N N s o ( Yn Yn ) n= Y o n, (3) MPE = N N n= Y s n Y o Y n o n, (4) were N is te number of observations, simulated traffic count. s Y n is te estimated observation and Y o n is te corresponding 5

10 Transport and Telecommunication Vol. 4, No, 03 Figure summarizes te RMSN results for all four intervals of te 5 scenarios tested in tis case study, considering also te tree alternative options used for te calculation of te Kalman Gain matrix of te Limiting EKF algoritm. Figure presents te corresponding MPE results. Te no-calibration case is presented to indicate te starting point, wile te EKF version presents a reference performance of te computationally intensive exact algoritm. Te tree Limiting EKF variants (depending on te way tat te H matrix was pre-computed) are also presented. Wile te various scenarios are not really connected and terefore a bar plot would be a more appropriate grapical representation lines are used in tis figure to provide a visually clearer presentation of te various cases. Te results suggest tat, as expected, te EKF algoritm performs well, and it manages to provide a satisfactory correction of te initial erroneous values of te parameters. Tese trials demonstrate te utility of te EKF algoritm in te OD estimation and prediction problems. It is also interesting to note tat, altoug te Limiting EKF algoritm eliminates te computationally intensive steps of te EKF algoritm, it does not lack significantly in terms of accuracy in its estimation. Te results of te Limiting EKF algoritm for all scenarios presented on Figures and sow tat in some cases te simpler Limiting EKF algoritm provides accuracy comparable to tat of te EKF algoritm. In some scenarios (suc as Scenario, in wic te true demand as been perturbed by +0% wile te true surveillance measurements ave been perturbed by -0%, tus misleading te calibration process) te EKF does not provide an improvement (and even provides a deterioration from te starting point). Te Limiting EKF, on te oter and, wic uses an average of formerly calibrated H matrices, manages to clearly outperform te EKF in tis case. Overall, it is noted tat te computation of te H matrix used in te Limiting EKF algoritm plays an important role in te accuracy of te estimation results. More specifically, te results indicate tat te computation of te H matrix used in LimEKFa (average matrix for eac time interval for te scenarios wit varying OD demand ±0%) and LimEKFc (average matrix across all time intervals for te scenarios wit varying OD demand ±0%) performed better tan in te case of te LimEKFb (average matrix for eac time interval for te scenarios wit varying OD demand ±0%). Actually, te latter case seemed to perform rater poorly in a few situations. Te approac of te average values tat was introduced in tis researc manages to eliminate te noise in te simulations, witout producing adverse estimation results. Te exact values of te results are presented in a table in Appendix A. Interval Interval RMSN (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c RMSN (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c Scenario Scenario Interval 3 Interval 4 RMSN (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c RMSN (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c Scenario Scenario Figure. Overview of te calibration results of te syntetic network based on te RMSN statistic 6

11 Transport and Telecommunication Vol. 4, No, 03 Interval Interval MPE (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c MPE (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c Scenario Scenario Interval 3 Interval 4 MPE (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c MPE (unitless) Mezzo EKF LimEKF_a LimEKF_b LimEKF_c Scenario Scenario Figure. Overview of te calibration results of te syntetic network based on te MPE statistic Case Study : Real-World, Large-Scale e Network Having establised tat te t Limiting EKF can provide favourably results (often comparable to te EKF) in a small, syntetic network, te next step of tis researc was w to evaluate te scalability of te experiments and to assess tee transferability and applicability of te Limiting L EKFF algoritm to a real- area; for a description of te network and tee calibration process c.f. Burgout et al. [4]) of Stockolm city was used. Tis area is onee of te most igly populated areas in i Scandinavia. Te Södermalm network consists of 46 OD pairs, 0 links, and 577 nodes, and includes te urbann arterial network of world, large-scale network. For tis t task, te already validated network of o te central district (Södermalm te area as well as te Södra Länken motorway, located at te sout endd of Stockolm. Te total number of possible routes in te network is 4,49. Bot Södra Länken motorway and its connection to te west (E4 motorway), experience congestion during te morning and afternoon peak ours, since tese are te major commuter routes to and a from te central district. A scematic of te Södermalm network as modelled witin Mezzo is presented on Figuree 3. Figure 3. Södermalm network modelled in Mezzoo 7

12 Transport and Telecommunication Vol. 4, No, 03 Tis scenario was also simulated for eigt 5-min intervals (i.e. a total duration of two ours), considering te first our as te warm-up period. In tis set of experiments two levels of te OD demand were evaluated (set at -0% and -0% of te original demand), and perfect knowledge of traffic counts was available across a small sample of te network links. For determining wic traffic count measurements to use in te limiting EKF algoritm, a simple sampling tecnique of selecting one every ten (non-zero) links was undertaken. Tis resulted to a total of 85 measurements. Similar to te syntetic network, it was furter assumed tat te covariance matrices Q and R are identity matrices. In addition, a 5% value was assumed for te linear delta factor, indicating te percentage of te demand to be considered during te linearization step. As an intermediate step, te EKF algoritm was run for eac scenario and te computed H matrices for te four intervals were used to calculate te average H matrix to be used in te Kalman gain matrix G of te Limiting EKF algoritm. Te algoritm was evaluated in terms of its ability to estimate te traffic flow conditions at te measurement locations, troug te RMSN and MPE statistics (as in case study ). Te results of te evaluation of te two scenarios are presented in Table. Table. Summary results of te Limiting EKL algoritm (Södermalm network) Mezzo LimEKF Improvement RMSN MPE RMSN MPE RMSN MPE S. (-0%) Int. 4.% -3.6%.7% -4.4% 47.3% 67.6% Int. 7.7% -9.8% 4.3% -6.% 48.4% 68.7% Int % -7.8% 3.5% -3.6% 54.% 79.8% Int % -3.9% 6.0% -3.4% 39.4% 75.5% S. (-0%) Int. 3.% -7.%.8%.0% 9.9% 85.9% Int. 5.0% -9.7% 3.5% -8.5% 0.0%.4% Int % -6.6% 5.7% -5.5% 6.0% 6.7% Int. 4 6.% -8.% 5.4% -5.7% 4.9% 9.6% Based on te results sown in Table, te Limiting EKF algoritm provides significantly lower (i.e. better; between 40% and 80%) RMSN and MPE statistics compared to te base scenario, wen te OD demand is -0%. For smaller variation of te OD demand te algoritm still appears to perform as expected, altoug it is not as efficient in minimizing te difference between te simulated and observed conditions. Still, te improvement is between 5% and 30%, wit a single value exceeding 85% improvement). In conjunction wit te results of te syntetic network, it can be inferred tat te LimEKF algoritm performs well wit regards to correcting te erroneous values of te OD demands, tus providing more accurate estimation of te prevailing conditions, especially wen te discrepancies are significant. Scatterplots of observed vs. simulated traffic counts at te available network links using te results from te Limiting EKF for OD estimation for te two scenarios discussed above, are presented in Figure 4. Besides a sligt underestimation of larger traffic counts, te results are fairly consistent wit te 45 degree straigt line tat represents te ideal results. It is noted tat suc an underestimation is expected, as te starting demand is significantly lower tan te true demand. Scenario Scenario simulated traffic counts (ve/) Interval Interval Interval 3 Interval 4 simulated traffic counts (ve/) Interval Interval Interval 3 Interval observed traffic counts (ve/) observed traffic counts (ve/) Figure 4. Scatterplots of observed and simulated traffic counts at te Södermalm network for te Limiting EKF algoritm 8

13 Transport and Telecommunication Vol. 4, No, 03 Conclusions Te results of te EKF algoritm (Algoritm ) using a syntetic network sowed tat te algoritm provides acceptable corrections and minimizes te discrepancy between te simulated and observed traffic conditions. To overcome te computational burden of te linearization step during te on-line calibration, a variation of te EKF algoritm, te Limiting EKF algoritm, was also evaluated. Te Limiting EKF algoritm computes te Kalman gain matrix off-line, taking into account precomputed gain matrices. Experiments in te syntetic network sowed tat, te Limiting EKF algoritm provides accuracy comparable to tat of te best algoritm (EKF), wile providing substantial improvement in computational efficiency. As it was sown in Figures and, in most cases tese results were consistent among te tree different approaces for computing te Kalman gain matrix. Te increased robustness of te proposed algoritm can be explained by te fact tat te average of various precomputed Kalman gain matrices reduces te noise tat is introduced in te computation of te individual Kalman gain matrices. Additional experiments undertaken in an existing large-scale network furter validate tese findings. Te efficiency of te Limiting EKF algoritm increases wit increased variation of te OD demands, wereas it s estimating and predictive power diminises for low OD demand variations. Future (ongoing) researc includes trying additional combinations of Kalman gain matrices for te computation of te Limiting Kalman gain matrix, as well as testing te generalization and scalability of te obtained results; for example, weter te results would old wen using oter traffic simulators and wen applying te approac on larger, more complex networks. In addition, ongoing researc performed using te Södermalm network evaluates te robustness of te LimEKF algoritm and te properties of te Kalman gain matrix, considering various structures of te error covariance matrices Q and R, as well as te impact of te number of available traffic counts in te network. Acknowledgments Tis work was supported in part by NTUA Basic Researc Grant 65/840 and benefited from te participation of te first two autors in te EU COST Action TU0903 MULTITUDE. Te autors would like to tank Dr. Wilco Burgout for elp and support relating to te interfacing of te mezzo simulator witin te developed framework. References. Balakrisna, R., Ben-Akiva, M. and H. N. Koutsopoulos (007). Off-line Calibration of Dynamic Traffic Assignment: Simultaneous Demand and Supply Estimation. Transportation Researc Record, No. 003, Antoniou, C., Balakrisna, R., Koutsopoulos, H. N. and M. Ben-Akiva (0). Calibration metods for simulation-based dynamic traffic assignment systems. International Journal of Modeling and Simulation, 3(3), Tavana, H. and H. Mamassani (000). Estimation and application of dynamic speed-density relations by using transfer function models. Transportation Researc Record, No. 70, Huyn, N., Mamassani, H. and H. Tavana (00). Adaptive speed estimation using transfer function models for real-time dynamic traffic assignment operation. In Proceedings of te 8 st Annual Meeting of te Transportation Researc Board, January 3 7, 00 (cd-rom). Wasington, D.C.: Transportation Researc Board. 5. Qin, X. and H. Mamassani (004). Adaptive calibration of dynamic speed-density relations for online network traffic estimation and prediction applications. In Proceedings of te 83 rd Annual Meeting of te Transportation Researc Board, January 5, 004 (cd-rom). Wasington, D.C.: Transportation Researc Board. 6. Antoniou, C., Ben-Akiva, M. and H. N. Koutsopoulos (005). On line calibration of traffic prediction models. Transportation Researc Record: Journal of te Transportation Researc Board, No. 934, Wang, Y. and M. Papageorgiou (005). Real time freeway traffic state estimation based on Extended Kalman Filter: A general approac. Transportation Researc Part B, 39, Wang, Y., Papageorgiou, M. and A. Messmer (May 007). Real time freeway traffic state estimation based on Extended Kalman Filter: A case study. Transportation Science, 4,

14 Transport and Telecommunication Vol. 4, No, Wang, Y., Papageorgiou, M., Messmer, A., Coppola, P., Tzimitsi, A. and A. Nuzzolo (009). An adaptive freeway traffic state estimator. Automatica, 45(), Asok, K. and M. Ben-Akiva (993). Dynamic O-D matrix estimation and prediction for real-time traffic management systems. In C. Daganzo (Ed.), Transportation and Traffic Teory, (pp ). Amsterdam, Te Neterlands: Elsevier Science Publising.. Asok, K. and M. Ben-Akiva (000). Alternative approaces for real-time estimation and prediction of time-dependent origin-destination flows. Transportation Science, 34(), 36.. Antoniou, C. (997). Demand simulation for dynamic traffic assignment. Master s tesis, Massacusetts Institute of Tecnology. 3. Ben-Akiva, M., Bierlaire, M., Koutsopoulos, H. N. and R. Misalani (00). Gendreau, M. and Marcotte, P. (Eds), Transportation and network analysis: current trends, capter Real-time simulation of traffic demand-supply interactions witin DynaMIT, (pp. 9 36). Dordrect, te Neterlands: Kluwer Academic Publisers. Miscellenea in onor of Micael Florian. 4. Ben-Akiva, M., Koutsopoulos, H.N., Antoniou, C. and R. Balakrisna (00). Traffic Simulation wit DynaMIT. In J. Barcelo (Ed.), Fundamentals of traffic simulation, (pp ). New York: Springer. 5. Bierlaire, M. and F. Crittin (004). An efficient algoritm for real time estimation and prediction of dynamic OD tables. Operations Researc, 5(), Zou, X. and H.S. Mamassani (004). A structural state space model for real time origin destination demand estimation and prediction in a day to day updating framework. In Proceedings of te 83 rd Annual Meeting of te Transportation Researc Board, January 5, 004 (cd-rom). Wasington, D.C.: Transportation Researc Board. 7. Antoniou, C. (September 004) On line Calibration for Dynamic Traffic Assignment. PD dissertation, Massacusetts Institute of Tecnology, Cambridge. 8. Antoniou, C., Ben-Akiva, M., and H. N. Koutsopoulos (007). Non linear Kalman filtering algoritms for on line calibration of dynamic traffic assignment models. IEEE Transactions on Intelligent Transportation Systems, 8(4), Asok, K. (996) Estimation and Prediction of time-dependent Origin-Destination Flows. PD dissertation, Massacusetts Institute of Tecnology. 0. Kalman, R. E. (960). A new approac to linear filtering and prediction problems. Journal of Basic Engineering (ASME), 8D, Cui, C. K. and G. Cen (999). Kalman Filtering wit Real-Time Applications. Heidelberg: Springer-Verlag.. Burgout, W. (004) Hybrid Mesoscopic-Microscopic Traffic Simulation. PD Tesis, Royal Institute of Tecnology (KTH). 3. Burgout, W., Koutsopoulos, H., Andreasson, I. (005). Hybrid mesoscopic-microscopic traffic simulation. Transportation Researc Record, No.934, Burgout, W, Koutsopoulos, H. N. and I. Andreasson, (00). Incident Management and Traffic Information Tools and Metods for Simulation-Based Traffic Prediction. Transportation Researc Record, No. 6,

15 Transport and Telecommunication Vol. 4, No, 03 Appendix A Table A.. Summary results of te EKF and Limiting EKF algoritm (/) Mezzo EKF LimEKF (a) LimEKF (b) LimEKF (c) RMSN MPE RMSN MPE RMSN MPE RMSN MPE RMSN MPE S. Int. 0. 0% 0. 0% 0% 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% Int. 0. 0% 0. 0% 0% 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% Int % 0. 0% 0% 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% Int % 0. 0% 0% 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% 0. 0% S. Int. 0. 0% 0. 0%. 8% -. 5% 3. % -3. 0%. 9% -. 7% 3. 6% -3. 4% Int. 0. 0% 0. 0% 3. 4% -. 7% 5. 4% -5. 0% 3. 9% -3. 0% 4. 0% -3. 5% Int % 0. 0% 3. 6% -3. 5% 3. 5% -3. 4% 3. % -. 9% 4. 0% -4. 0% Int % 0. 0% 3. 0% -. 8% 3. 4% -3. 0% 6. 9% -6. 9% 3. 3% -3. 3% S. 3 Int. 0. 0% 0. 0%. 9%. 9% 3. 5% 3. 4% 3. % 3. % 4. % 4. % Int. 0. 0% 0. 0% 3. 4% 3. 4% 5. 8% 5. 8% 4. 9% 4. 7% 4. % 4. % Int % 0. 0% 4. % 3. 9% 3. 9% 3. 8% 3. %. 8% 4. 3% 3. 9% Int % 0. 0% 3. 9% 3. 9% 3. 7% 3. 7% 9. 3% 9. % 3. 9% 3. 9% S. 4 Int. 9. 0% 9. 0% 3. % 3. 0% 3. % 3. % 3. 0% 3. 0% 3. 6% 3. 6% Int. 0. 5% 0. 3% 4. 0% 3. 9% 4. 0% 3. 8% 4. % 4. % 3. 9% 3. 7% Int % 9. %. 7%. 7% 3. 6% 3. 3% 3. 5% 3. 5% 3. 5% 3. 5% Int % 0. 4% 3. 4% 3. % 3. 7% 3. 4% 7. 5% 7. 3% 4. % 4. 0% S. 5 Int. 9. 0% 9. 0%. 6% -0. 7%. 4% -0. 5%. 3% -0. 3%. 7% -0. 5% Int. 0. 5% 0. 3%. 6% -0. 5%. 8% -0. %. % 0. 8%. 3% 0. 0% Int % 9. %. 7% -0. 4%. 0% -0. 3%. % -0. 4%. 7% -0. 4% Int % 0. 4% 0. % 0. % 0. 3% -0. 3% 0. 7% 0. % 0. 4% -0. % S. 6 Int. 9. 0% 9. 0% 6. 9% 6. 9% 6. 7% 6. 7% 6. % 6. % 7. 9% 7. 9% Int. 0. 5% 0. 3% 8. 6% 8. 5% 0. 8% 0. 8% 9. 0% 9. 0% 8. 9% 8. 8% Int % 9. % 8. 6% 8. 6% 8. 4% 8. 4% 6. % 5. 8% 8. 7% 8. 7% Int % 0. 4% 8. 0% 8. 0% 7. 9% 7. 8% 8. 3% 8. % 8. 6% 8. 6% S. 7 Int. 9. % -9. %. 8% -. 8%. 8% -. 8%. 3% -. 4% 3. 3% -3. 3% Int. 0. % -0. % 3. 6% -3. 6% 4. 9% -4. 9% 3. 6% -3. 6% 3. 9% -3. 9% Int % -9. 8% 4. % -3. 9% 4. 0% -3. 6% 3. % -. 5% 4. % -3. 7% Int % -0. 0% 4. 9% -4. 8% 3. 6% -3. 4% 7. 7% -7. 6% 4. % -3. 8%

16 Transport and Telecommunication Vol. 4, No, 03 Table A.. Summary results of te EKF and Limiting EKF algoritm (/) Mezzo EKF LimEKF (a) LimEKF (b) LimEKF (c) RMSN MPE RMSN MPE RMSN MPE RMSN MPE RMSN MPE S. 8 Int. 9. % -9. % 5. 5% -5. 4% 5. 6% -5. 6% 4. 7% -4. 7% 6. 3% -6. % Int. 0. % -0. % 6. 6% -6. 3% 9. % -8. 9% 7. 5% -7. % 7. 0% -6. 8% Int % -9. 8% 6. 0% -5. 9% 5. 9% -5. 9% 4. 8% -4. 7% 6. 9% -6. 9% Int % -0. 0% 5. 9% -5. 9% 6. 0% -6. 0% 4. 6% -4. 5% 7. 0% -7. 0% S. 9 Int. 9. % -9. % 0. 7% 0. 3% 0. 7% 0. 3% 0. 7% 0. 5% 0. 68% 0. 46% Int. 0. % -0. % 0. 5% 0. 0% 0. 8% 0. 0% 0. 6% -0. 3% 0. 5% 0. 0% Int % -9. 8% 3. 8% 0. 9% 3. 4% 0. 3% 3. 9% 0. 5% 3. 58% 0. 8% Int % -0. 0%. 9% 0. 0%. 9% -0. 4%. 4% 0. %. 93% % S. 0 Int. 7.4% 7.3% 5.6% 5.5% 6.% 6.% 5.9% 5.8% 7.% 7.% Int. 9.% 8.9% 5.0% 4.7% 0.0% 9.9% 8.4% 8.3% 8.4% 8.3% Int % 4.%.3%.% 0.4% 0.3% 9.% 8.9% 0.4% 0.3% Int % 0.6% 0.% 0.0% 7.% 7.0% 6.6% 6.6% 8.4% 8.4% S. Int. 7.4% 7.3%.7%.9%.9%.% 3.%.3%.9%.3% Int. 9.% 8.9% 9.% 8.4% 5.3% 4.3% 6.0% 5.3% 6.0% 5.% Int % 4.% 3.%.9%.9%.% 0.7% 0.3%.8% 0.8% Int % 0.6% 4.4% 3.9% 3.3% 3.3% 9.3% 8.8% 4.7% 4.4% S. Int. 7.4% 7.3% 9.5% 9.6% 0.7% 0.8% 9.7% 9.7%.9%.0% Int. 9.% 8.9% 8.% 8.% 7.% 7.% 3.7% 3.7%.9%.9% Int % 4.% 3.9% 3.6% 5.5% 5.4% 3.9% 4.0% 6.7% 6.6% Int % 0.6% 40.8% 40.8%.3%.3% 7.% 7.%.5%.5% S. 3 Int. 8.7% -8.7% 5.0% -5.0% 5.6% -5.6% 4.8% -4.8% 6.% -6.% Int. 9.5% -9.5% 6.% -6.% 9.8% -9.8% 6.7% -6.6% 7.6% -7.7% Int % -0.% 8.5% -7.8% 7.4% -6.6% 6.0% -4.7% 8.3% -7.6% Int. 4 0.% -0.0% 8.9% -8.8% 6.6% -6.4% 6.% -6.% 7.8% -7.6% S. 4 Int. 8.7% -8.7% 7.0% -7.0% 8.0% -7.9% 6.7% -6.7% 8.7% -8.7% Int. 9.5% -9.5% 9.% -8.9% 3.6% -3.5% 9.8% -9.6% 0.% -0.% Int % -0.% 9.4% -9.3% 9.% -8.9% 7.3% -7.0% 0.5% -0.4% Int. 4 0.% -0.0% 0.5% -0.5% 8.5% -8.4%.8% -.7% 0.3% -0.% S. 5 Int. 8.7% -8.7%.5% -.5%.9% -.9%.5% -.5% 3.% -3.% Int. 9.5% -9.5% 4.4% -4.3% 5.6% -5.5% 3.6% -3.7% 4.0% -4.0% Int % -0.% 5.6% -3.8% 5.6% -4.0% 3.8% -.% 5.4% -4.0% Int. 4 0.% -0.0% 5.% -4.3% 3.6% -3.4% 9.3% -9.% 4.8% -4.3%

17 Transport and Telecommunication Vol. 4, No, 03 Transport and Telecommunication, 03, Volume 4, No, 3 9 Transport and Telecommunication Institute, Lomonosova, Riga, LV-09, Latvia DOI 0.478/ttj TELEMATICS BUSINESS AND MANAGEMENT IN BAHRAIN MARKET Marwan Raje Hussain, Abduljalil Zainal, Wael Moamed Elmedany 3, Moamed Waleed Fakr 4 Telematics Department (consultant), Spatial Tecnology Solutions (STS),University of Barain P.O Box 3038, Manama, Kingdom of Barain marwan.raje@sts-int.net CEO, Spatial Tecnology Solutions (STS), University of Barain Manama, Kingdom of Barain abduljalil.zainal@sts-int.net 3 Computer Engineering Department, University of Barain P.O. Box 3038, Manama, Kingdom of Barain welmedany@uob.edu.b 4 Electronic Engineering Department, University of Barain Manama, Kingdom of Barain waleedfakr@yaoo.com Te aim of tis paper is to present and discuss issues related to te telematics business and management in Barain. Telematics is a part of Information and Communication Tecnologies (ICT), wic plays a major role in te economic development of countries. Telematics applications are becoming increasingly important in modern transportation to increase reliability, safety and security of bot veicles and drivers. Tere are many applications for telematics; some of tese applications are telematics for educations, telematics for ealt, and veicle telematics. One of te most important applications of telematics is veicle telematics, wic includes veicle tracking, fleet management, container tracking, trail tracking. In tis paper we are focusing on veicle telematics business and management in Barain market. Keywords: Telematics, ICT, Veicle Telematics, GPS, GSM, GPRS, navigation, fleet management. Introduction Information and Communication Tecnologies (ICTs) denote a wide range of services, applications and tecnologies, using various types of ardware and software, often running over telecommunication networks. Te importance of ICTs is te enabling function in facilitating enanced access to information and communication across large distances []. One of te most important components of ICT is te telematics, wic in turns includes many oter components and devices tat are used in different telecommunication applications []. Telematics is defined as an integrated use of telecommunications and informatics; it is about te sending, receiving and storing information via telecommunication devices (GSM/GPRS) along wit controlling of remote objects [3], [4], [5]. Telematics applications are becoming increasingly important for auto insurers in numerous countries as tey seek to control claim costs, enance pricing sopistication, improve profitability, and differentiate teir policyolder products and services [3]. Telematics nowadays plays a major role in te economic development of countries. It includes but not limited to Global Positioning System (GPS) tecnology integrated wit computers and mobile communications tecnology in automotive navigation systems [6], [7]. Wit te enormous market potential of te telematics industry and te rapid development of information tecnology, automotive telematics as attracted considerable attention for mobile computing and Intelligent Transport Systems (ITSs) [8]. Te telematics services, includes, navigation services, remote diagnostics, emergency calls and more [9]. Tere are many telematics applications examples around te world, many publications for telematics applications and developments of telematics in different courtiers are already online, as an example is study of te telematics industry in Korea and Cina [0]. In Barain, like in te oter Gulf Co-operation Council (GCC) Countries, tere is an urgent need to increase te number of telematics companies and at te same time to create favourable conditions for emerging private enterprises tat require better knowledge in economics, business, and company management. 3

18 Transport and Telecommunication Vol. 4, No, 03 Tis paper discusses te telematics business and management in Barain market, te most important applications of telematics in Barain is veicle telematics application and navigation using PNDs. Te paper organized as follows; after te introduction in section one, te importance of telematics and its contribution in economic development is presented in section two; section tree will discuss te telematics business in Korea and Cina; section four is discussing te telematics business in Barain Market; after tat we are going to present some of te available telematics applications in Barain in section five; finally, a conclusion is presented in section six.. Importance of Telematics and Its Contribution in Economic Development Te expansion of information and communication tecnologies (ICT) from ardware, software, Internet, telepone, content, information, application, support service [] and te growt in wireless connectivity as canged te way people live, work, and play. New knowledge and te use of new ICT tecnologies ave resulted in te creation of new products, services, and jobs, some of wic were unbelievable only a few decades ago. At te same time, trade agreements and te reduced costs of communications and transportation ave increased te flow of capital, goods, services, knowledge and jobs between countries, te result as been significant worldwide economic growt []. As a result, it as been noticed tat firm s using ICT as faster growt, iger productivity and faster employment growt. Moreover, ICT elped in growing individual economic opportunity and enabled people to enance teir knowledge and skills. Also, it elped individual to use teir income more wisely and manage teir own businesses efficiently []. Telematics is part of Information and communication tecnologies (ICT) wic plays a major role in te economic development of countries. As it as been mentioned in earlier sections tat telematics is about te tecnology of sending, receiving and storing information via telecommunication devices (GSM/GPRS) along wit controlling of remote objects/devices. Terefore, telematics as te following main caracteristics [0]: Timeliness, by providing users wit immediate access to information anytime. Mobility, by providing services anywere. Individuality, by providing services reflecting te user s geograpical location and demand. Convenience, by providing useful information and services. Tere are various key players aving important role in te Telematics industry wic contributes in economic development. Tey are basically classified into tree categories; software, ardware and network related. Among te software related players, tere is a player called te service provider wic provides te users wit te Internet or telematics services suc as web browsing, veicle location, ing, and advertising. Anoter software related player is called te contents provider wic provides te user wit contents and data. Basically, software related providers are operated by automakers, auto dealers, internet service providers, advertisement firms, digital map distribution firms, or several oter facilities [3]. Hardware related players are automakers, consumer s electronics firms, communication equipment firms, and in-veicle device suppliers wic provide te car navigation system tat is expected to be a communication device for telematics. In addition, vendors of mobile pones and PNDs (Portable Navigation Devices) will ave te potential to be ardware related players. Network related players provide network infrastructure suc as mobile communication infrastructure, wireless LAN, and ground station digital TV broadcasting, wic are essential tecnologies for telematics. Telecommunication companies, broadcasting companies, and IT solution vendors wic connect telematics to oter businesses can be listed as network related players [3]. Terefore, all tose players will support in providing te telematics services wic indeed will contribute in economic development. 3. Telematics in Korea and Cina Korea and Cina ave witnessed significant economical development in recent years, mainly as a result of teir tecnological development in te areas of ICT (information and communication tecnology) and semiconductor tecnology. In particular, te telematics industry, wic incorporates information tecnology (IT) into automobiles, as received increasing attention by users, owever, tere are many issues surrounding te telematics industry tat need to be addressed bot academically because te industry is still in its initial stages of te production life cycle and tere remain substantial development gaps between countries. On average, te global telematics industry as grown 37.8% per year since 006, and it is expected to reac USD 9.4 billion. Te telematics industry as been providing stakeolders wit, new opportunities for developing and strengtening teir value cain [0]. 4

19 Transport and Telecommunication Vol. 4, No, 03 In Korea, te demand for telematics as been increased by te government, fleet owners and individuals because of te countries complex road conditions and frequent traffic jams. Based on Korea s world-class, telecommunications infrastructure and mobile communications tecnology and te tecnical competitiveness of te automobile industry, te country s telematics market is expected to become an important part of te global telematics industry. In Cina te government selected te telematics industry as one of Cina s tree main growt engines and invested more tan RMB 0 billion to produce more tan 00 million cars wit telematics tecnology by 00. Since ten, many stakeolders, including tose in te automobile industry, ave regarded te telematics industry as an important symbol of advancement, innovation, and value added and tus, ave focused on developing and providing a wide range of telematics products and services. In recent years, Korea and Cina, taking advantage of teir geograpical proximity, ave engaged in bilateral negotiations and trade to expand teir telematics presence. 4. Telematics Market Development in Barain In Barain te demand for Veicle Telematics Management Systems (VTMS) as been increasing because fleet owners are looking for reducing teir costs, utilize teir fleet and maximize productivity. Moreover, due to te big jump in telecommunication infrastructures and mobile communication tecnology made it easy for businessman (specialized or non-specialized) to enter suc market in Barain and provide tis service. Veicle Telematics started in Barain in 006. Tere were two companies providing tis service for fleet owners in public and private sectors. One of tose companies started tis business by providing full system component (configuration, application, servers, GIS Barain map, website osting and system support) in-ouse troug its umble R&D section, except te Telematics units was purcased from outside suppliers in US, Korea and Cina. Te oter company was dealing wit tird party outside Barain wic provides full veicle Telematics system components wit very low cost and low quality. Terefore, tis company does not ave any team for development; simply buy and sell. Ten in 007 anoter company joined tis market and followed by anoter company in 009. After tat in 00 and 0 anoter tree companies entered te veicle Telematics business. Finally, in 0 anoter five companies added to te list of companies tat provide suc products and service in Barain Market (see Table I and Figure ). Table. Number of companies tat provide Veicle Telematics products and services in Barain since 006 to 0 Year # of Companies Figure. Number of companies tat provide Veicle Telematics products and services in Barain since 006 to 0 5

20 Transport and Telecommunication Vol. 4, No, 03 Te following (Table II and Figure ) demonstrate te growt potential of Barain veicle telematics market and te number of veicle telematics units mounted so far since 006 to 0. Table. Number of veicles wit mounted veicle telematics system in Barain since 006 to 0 Year # of Veicles wit Telematics system Figure. Number of veicles wit mounted veicle telematics system in Barain since 006 to 0 It is clear from te earlier graps and tables tat te number of companies entered te market of Telematics in 0 ave increased dramatically along wit te number of mounted veicle telematics units in same year. Tis big cange is due to te demand in modern transportation for veicle telematics to increase reliability, safety and security of bot veicles and drivers. Moreover, te political situation arise in Barain in 0 as caused to increase te demand of veicle telematics system too. As a result of te growt in Barain s population and te prosperous economy, te number of registered veicles in te kingdom as been growing at average rate of 6.7% between 005 and 0; te total number of veicles on Barain roads reaced ig of over 475 tousand at te end of 0 (see Table III and Figure 3) [4]. Tis means tat still tere is a big potential for te veicle telematics business to expand more in Barain market. Table 3. Number of Veicle Registered and in use (005 0) [4] Year # of Veicles Registered % of increase , , , , , , ,50 5 6

21 Transport and Telecommunication Vol. 4, No, 03 Figure 3. Number of Veicle Registered and in use (005 0) [4] 5. Available Telematics Applications in Barain Te most important applications of telematics in Barain are veicle telematics application and navigation. Veicle telematics is a way of monitoring te location, movements, status and beaviour of a veicle or fleet of veicles (suc as sedan cars, trucks, containers, motorcycles, etc ) remotely to enance te functionality, productivity and security of bot veicles and drivers [5]. SmartTrack in one of te famous veicle telematics application tat is available in Barain. Tis is te flagsip telematics product offered by Spatial Tecnology Solutions (STS) one of te newly establised ICT companies in Barain [6]. SmartTrack is a complete web-based fleet tracking and management solution, wic comes wit a wide array of standard and optional features to suite every customer needs. SmartTrack is built around proven ardware and software tecnologies from leading H&S tecnology providers in te world including Microsoft (for te back-end server), ESRI GIS for te mapping software, and leading GPS engine suppliers. Te tree main components of SmartTrack are te following: Veicle-mounted tracking unit (VTU): a black box integrating a GPS engine wit a GSM/GPRS mobile communicator to communicate wit te back-bone servers platform were te receiver application and databases are osted. SmartTrack fleet Application: a web-based application tat offers te users te ability to track teir veicles live, and to obtain istorical reports giving details of teir veicles movements and status at any given time, as sown on Figure 4 6. SmartTrack Hosting platform: Tis is te back-bone servers platform wic ost te applications and mapping engine of SmartTrack. Figure 4. STS Digital Map View of Veicle Tracking in Barain 7

22 Transport and Telecommunication Vol. 4, No, 03 Figure 5. Sample of istorical veicle points (in Trip Report) viewed on STS Barain map in SmartTrack website Figure 6. Smaple Montly Summary Report of Veicle Telematics (SmartTrack) in Barain Customers using suc tecnology are many and include emergency veicles (e.g. ambulance, electricity, water and fire brigade), delivery cars, construction veicles, taxis, waste collection and buses. Te benefits realized from using tis tecnology includes te following; improved driver/veicle management, reduced fuel consumption (ence reduced pollution), reduced maintenance cost, reduced accident & misuse, and improved customer delivery services and improved fleet efficiency. Finally, navigation is te second application used as one of te Telematics services provided in Barain. Typically, tis involves te use of Personal Navigation Devices (PND) also called Portable Navigation Devices wic are portable electronic devices wic combines a positioning capability (suc as Global Positioning Systems (GPS)) and navigation functions is muc convenient, safer and reliable solution for drivers to find teir destinations witout te need of aving a paper based maps [7]. All PNDs work essentially te same way. Tey store map data locally on te device, wic must be updated wit some frequency (quarterly or annually) and support routing algoritms (te fastest route or te sortest route to destination) [8]. Some newer PNDs provide some real-time weater and traffic data via a data connection over a cellular network [9]. 6. Conclusions In tis paper we ave presented and discussed issues related to telematics business and management in Barain market. Te paper also discussed telematics in Korea and Cina, as an example for telematics applications in oter countries. As an example of telematics business in Barain, te paper discussed te growt of veicle telematics products and services in Barain since 006 to 0. Te provide table and cart demonstrated te growt potential of Barain veicle telematics market and te number of veicle telematics units installed during te period of year 006 to 0. Te carts and tables demonstrated tat tere is an increase in te use of veicle telematics products in Barain during te last six years, as indicated also by a large increase in te number of service providers during te same period. 8

23 Transport and Telecommunication Vol. 4, No, 03 References. Siriginidi Subba Rao. (009). Acieving millennium development goals: Role of ICTS innovations in India, Telematics and Informatics, Ming-Ciao Cen, Jiann-Liang Cen, Teng-Wen Cang. (0). Android/OSGi-based veicular network management system, Computer Communications, Laurie, A. (0). Telematics-Te New Auto Insurance, Retrieved September 3, 0, from 4. GPRS Services (0). Retrieved October, 0, from ttp:// 5. GPRS: General Packet Radio Service (0). Retrieved October, 0, from ttp:// (-0-0) 6. Hossaina, E., Cowb, G., Leungc, Victor C. M., McLeoda, R., Mišic, J., Wongc, Vincent W. S., Yange, O. (00).Veicular Telematics over eterogeneous wireless networks: A survey, Computer Communications, ScienceDirect, 33, GPS Personal Tracking. (0). Retrieved October, 0, from ttp:// gps+personal+tracking.tml 8. Sirazia, F., Golamib, R., Higo, D. An. (009). Te impact of information and communication tecnology (ICT), education and regulation on economic freedom in Islamic Middle Eastern countries, Information & Management, ScienceDirect, 46, Lenflea, Sylvain, Midlerb, Cristope. (009). Te launc of innovative product-related services: Lessons from automotive telematics, Researc Policy, Young-Wook Song, Ji Dae Kim, Liangri Yu, Hyun Kyung Lee, and Hyung Seok Lee. (0). A Comparative Study of te Telematics Industry in Korea and Cina, Cungbuk National University, Sout Korea. Retrieved November 9, 0, from ttp:// jibc/articles.tm. Kramer, William J., Jenkins, Bet and Robert S. Katz. (007). Te Role of te Information and Communications Tecnology Sector in Expanding Economic Opportunity. Harvard: Te Fellows of Harvard College.. Kozma, Robert B. (November 005). Education Reform, and Economic Growt. Retrieved December 3, 0, from ftp://download.intel.co.jp/education/wsis/ict_education_reform_economic_growt.pdf 3. Noriito Simizu. (June 004). Analysis of Automotive Telematics Industry in Japan, Massacusetts Institute of Tecnology. Retrieved December 3, 0, from ttp://dspace.mit.edu/andle/7./ Veicle Registered in Use, New and Cancelled (004 0), Table 5., Transport and Communications, (0). Retrieved November 6, 0, from ttp:// SubDetailed.aspx?subcatid= UK Telematics Online. VEHICLE TRACKING & VEHICLE TELEMATICS explained. Retrieved September 0, 0, from ttp:// 6. SmattTrack (Fleet Management System). Retrieved October, 3, 0, from ttp://smarttracksts.com/smarttrackv 7. Dispatc and Navigation Garmin to Contigo. (0). Retrieved October, 8, 0, from ttp:// eet.pdf 8. Personal Navigation Devices. Retrieved November 4, 0, from ttp:// index.pp/personal_navigation_devices 9. Personal Navigation Devices. Retrieved November 4, 0, from ttp://personalgps.wordpress.com/ applications/personal-navigation-devices/ 9

24 Transport and Telecommunication Vol. 4, No, 03 Transport and Telecommunication, 03, Volume 4, No, 0 8 Transport and Telecommunication Institute, Lomonosova, Riga, LV-09, Latvia DOI 0.478/ttj PLANNING AND FINANCING SCHEMES LINKED TO THE DECISION-MAKING FOR THE INTERCONNECTION OF LONG-SHORT DISTANCE TRANSPORT Eftiia Natanail, Giannis Adamos University of Tessaly, Department of Civil Engineering Pedion Areos, Volos, Greece P.: Fax: enat@ut.gr University of Tessaly, Department of Civil Engineering Pedion Areos, Volos, Greece P.: Fax: giadamos@civ.ut.gr Te development of interfaces between long and sort distance transport networks is associated wit te general processes planned by central governmental bodies; and te degree of teir efficiency depends eavily on te successful cooperation among te involved planning stakeolders at various territorial levels. At te same time, no matter ow efficient te planning process may be te lack of te appropriate financial planning and te coordination of te involved stakeolders for finding te required resources, affects te development of successful long and sort-distance interfaces. Towards tis direction, te aim of te present paper is to investigate te planning and financing scemes developed in longsort interconnection in representative European countries, and assess ow tese processes are linked to te decision-making processes at a local, regional, national and European level. In addition, obstacles and barriers in te relative processes are indicated and metods for teir alleviation are proposed. Keywords: modal interconnection, terminal interfaces, decision-making, planning, financing, questionnaire survey Introduction Transport as gradually become one of te basic fields of concern of te European policy, wit te Wite Paper European Transport Policy for 00 : Time to decide [], representing te first compreensive framework for te acievement of a sound European transport policy, and stating tat te common transport policy as to be part of an overall strategy integrating sustainable development, including land use-planning policy, budgetary and fiscal policy, etc. []. In 006, te Mid-review of te Wite Paper and te Communication A Sustainable Future for Transport: Towards an Integrated, Tecnology Led and User Friendly System [3], argued for a olistic approac to transport policy, taking into consideration tat mutually complementary action is needed at national, regional and local levels of government, aiming at integrating investment needs, regulations, differentiated solutions, tecnological innovation and infrastructure development. Te recently publised (0) by te European Commission Wite Paper Roadmap to a Single European Transport Area Towards a competitive and resource efficient transport system [4] sets objectives and actions for te development of a more competitive and sustainable transport system till 050, and includes initiatives for te elimination of significant barriers tat obstruct te improvement of key sectors of transport, like infrastructure, investments and armonization of legal frameworks. Assessing te above policies and initiatives of te European Commission (EC), it can be stated tat transportation is in an era of cange. For te acievement of te objectives of te EC, sustainability in bot passenger and freigt transportation is required, and issues suc as te integration of planning and financing processes, te coerent cooperation among local, regional, national and European stakeolders and te integration of efficient interfaces between sort and long distance networks for te promotion of intermodality and limitation of road transport sould be addressed. Especially focusing on intermodality, te development of interfaces between long and sort distance transport networks is associated wit te general processes planned by central governmental bodies and te degree of teir efficiency depends eavily on te successful cooperation among te involved planning stakeolders at various territorial levels. Issues suc as environmental protection, energy conservation, modal split and competitiveness, improvement of accessibility and regulatory restrictions, affect te planning process, as well as te range and implementation of efficient interfaces. 0

25 Transport and Telecommunication Vol. 4, No, 03 At te same time, no matter ow efficient or integrated te planning process may be te lack of te appropriate financial planning and te coordination of te involved stakeolders for te finding of te required resources, affects te development of successful long and sort-distance interfaces. Te need for integrating te cooperation between te involved agencies and autorities, and establising sensible financing scemes tat improve te efficiency and flexibility of te processes, is crucial. Towards tis direction, and based on te researc and work conducted till now in te framework of te FP7 project CLOSER (ttp:// te aim of te present paper is to investigate te planning and financing scemes developed in long-sort interconnection in representative European countries, Spain, France, Norway, Finland, Czec Republic and Greece, and assess ow tese processes are linked to te decision making processes at a local, regional, national and European level. In addition, obstacles and barriers in te relative processes are indicated and metods for teir alleviation are proposed. For te acievement of te above aim, a survey was conducted, in order to collect data regarding te decision-making, planning and financing processes developed in eac country, as well as at specific terminals involving long and sort-distance interfaces.. Metod Te metodology developed in te present paper includes review and analysis of te planning and financing processes developed in representative European countries, tus, Spain, France, Norway, Finland, Czec Republic and Greece, and refers to te interconnection of long and sort distance transport networks. In addition, data was collected regarding te planning and financing scemes developed at specific terminals involving long and sort interfaces. In te first case, a template was used, including key points addressing te planning and financing procedures in eac European country. Indicative aspects tat were addressed wen collecting information about te planning processes are te following [5]: Type of transportation plans developed, i.e. national, regional and local. Documentation of te transportation plans (i.e. objectives, responsibilities, involved actors, etc.). Public participation (if any) in te planning process. Regulatory framework. Consideration of environmental protection, energy consumption, safety and security, and social conscience. Assessment of relative barriers. Planning orizon, monitoring and assessment. Similarly, te financing scemes developed in te above countries were identified, separated into two stages, tus, procurement/development, and management/operation. In te second case, wen investigating te planning and financing processes at specific European terminals, a questionnaire survey was conducted in order to collect te appropriate information. Te questionnaire was filled in by te involved stakeolders during a face-to-face interview, and was structured into four parts. In te first part, information about te stakeolder and te terminal was collected, wile te second and tird parts included questions about te planning (i.e. responsible actors/stakeolders, public participation procedure, etc.), and financing (responsible actors/stakeolders, funding type, etc.) processes, respectively. Lastly, te fourt part of te questionnaire regarded te identification of barriers and te suggestion of metods for teir alleviation.. Illustration of Decision-Making Process in Europe Decision-Making (D-M) can be considered as te process, under te scope of wic, te responsible bodies coose te best alternative for te planning, financing, time sceduling and implementation of a project, te content of a procedure, etc. Focusing on transportation, te best decision (solution) is usually te outcome of an integrated tecnical and economical approac and evaluation. Due to te wide range of bot public and private bodies tat participate, and te absence of a coerent legal framework applied in all transportation modes or possible combinations of tem [6], te D-M in transport seems to be a rater complex process, and an ideal ; tus, widely applicable framework is still under consideration. A Decision Makers Guidebook DMG [7], was produced in order to provide a metodology for D-M and is addressed to all categories of bodies involved in D-M processes, e.g., politicians, stakeolders, individual citizens, etc. According to te DMG, tree approaces ave been identified in Europe:

26 Transport and Telecommunication Vol. 4, No, 03 te vision led approaces generated by an individual politician tat visualizes specific future plans for te development of is/er municipality, region, etc., te plan-led approaces, wic include a definition of objectives and an assessment of potential gaps or problems, before te establisment of a strategy and making of a (final) decision, and te consensus-led approaces tat foresee relevant discussions among stakeolders for te development of te best strategy [8]. In addition, investigating, for example, te involvement of relevant public and private bodies in te urban transport planning, a diverse of stakeolders is observed, wic can be generally beneficial to tis stage of planning, altoug, te acievement of te participation of different interested groups tat influence plans tat are finally developed, remains difficult [9]. Te frequency, te quality and te establisment or not of an integrated framework (e.g., periodic meetings, use of Internet, etc.), under wic te previously mentioned discussions among stakeolders take place, before te final D-M, form te bases for te substantial (or not) public participation in te D-M processes. Te DMG [8], defines five different levels of public participation; tus, provision of information, consultation, deciding togeter, acting togeter and supporting independent stakeolder groups [9]. Neverteless, te public participation, even if wen it is well designed and despite te increased demand of people to ave an active role in te D-M processes, cannot assure te success of any transportation project [0]. Tere are still, many problems and difficulties tat ave to be encountered for te provision of efficient transportation. Te CLOSER project, introduced a coerent decision-making (D-M) framework for te sort-long distance interconnection, taking into consideration te wide range of te public and private bodies tat are involved in te D-M processes. More specifically, it was defined tat te strategic framework for sortlong distance interfaces, furter specified as an integrated plan, is illustrated by planning and policy stakeolders, wic are eiter considered separately or in accordance to te development of te needed infrastructure, and te system operation. Te users, also, play a catalytic role and tey affect te D-M process, at any level, i.e. strategic, tactical or operational [5]. Te findings of te survey conducted in te framework of CLOSER, regarding te planning and financing scemes tat are developed in long-sort interconnection in European countries and specific terminals, are presented in te following paragraps. 3. Planning and Financing Processes in European Countries Spain In 005, te Plan Estratégico de Infraestructuras y Transporte (Strategic Infrastructures and Transport Plan PEIT) was approved, dealing wit te planning of all actions in te field of infrastructure and transport, and representing te largest ever drive to provide infrastructure in Spain (ttp://peit.cedex.es/). Te Spanis Strategy of Urban Sustainable Mobility (EEMS) was approved in 009 and was developed by te Ministry of Environment and te Ministry of Public Works and Transport, aiming at te establisment of a framework to orientate and provide coerence to te set of sectorial policies tat are involved in te consecution of a sustainable mobility. In addition, te Spanis Strategy of Climatic Cange and Clean Energy was created to accomplis te Spanis commitments regarding climatic cange and to promote te use of clean energies wit a focus on welfare, economic development and environmental protection (ttp:// documentation_cc.estrategia_cc/pdf/cle_ene_pla_urg_mea.pdf). In terms of intermodality, altoug tis topic was not considered as a relevant issue in te national transport planning strategy, te last five years te PEIT as begun to recognize te importance of intermodality, and considered te need for integrating transport services and improving te collaboration among te different stakeolders, wit te expectation to tackle some issues tat ave traditionally indered intermodal integration up to tis moment at national level. Te PEIT pursuits te promotion of intermodality bot for freigt and passenger transport. In te first case, an Intermodal Goods Plan was defined dealing wit issues like te structure of an intermodal network in Spain, te national and international intermodal corridors, te port, rail and airport intermodality, etc. In te case of passenger transport, PEIT developed an Intermodal Passenger Plan, wit te aim to establis te basis for te intermodal passenger system, including timetable coordination, pysical integration of te transport modes, te review of te framework of concessions for road transport of passengers, te introduction of systems to stimulate competition and quality, and te promotion of suttle bus services between population centres of a certain size and airports and stations wit ig-performance services, etc. (ttp://peit.cedex.es/). Lastly, focusing on financing, strategies included in te PEIT, refer to te National budget, wic sould remain te main source of te financing transport infrastructures and te promotion of te private section participation troug Public-Private Partnersips.

27 Transport and Telecommunication Vol. 4, No, 03 France Te transportation planning system in France is developed under tree levels, tus, national, regional and local planning, and te associated financial scemes sould obey to te overall negotiation tat leads to te planning processes. More specifically, at a national level, te National Sceme for Infrastructure syntesizes national priorities and investment priorities of te regions, troug a procedure of contractualisation among te central state and te eac region. At te regional scale, following te first acts of Decentralization (98), te linking of transport and spatial planning was solely acieved troug te Contrats de Plan Etat Région (Territorial Coerence Scemes SPER). Tese scemes, lasting 5 to 7 years, set out decisions made in terms of transport facilities, wit te central government s decision being dominant (ttp://enseignementsup-recerce.gouv.fr/cid5758/les-contrats-de-projets-etat-region-c.p.e.r tml). Tis strategic plan foresees tat te connections between transport and land planning are explicitly defined at regional level, but te regions lack some of te levers needed to ensure coordination of spatial policies. On te oter and, te municipalities remain in control of land use planning, and tis causes unbalance to te coerence of te different levels of autorities, terefore te Scémas Régionaux d Aménagement et de Développement du Territoire (Patterns of Regional Development and Territorial Development SRADT) must oversee te coordination of te various Scémas de Coérence Territoriale (SCOT), but te latter do not ave to follow regional coices, and links between te Urban Transport Plans (PDUs) and te SRADTs do not exist. Regarding PDUs, tese sould be compatible wit te SPERs tat define te general orientations for te urban territory. In tis context, recent Frenc legislation as encouraged planning wic links transport and spatial planning at two levels: on one and, te conurbations wit te autorities responsible for urban transport carged wit creating PDUs and te municipal associations responsible for SCOT; on te oter and, te regions, wic draw up Scémas Regionaux de Transport (Regional transport scemes SRTs) witin te framework SRADTs. Norway Te Norwegian National Transport Plan (NNTP) for te period is te central document for te transport sector in Norway, produced every four years, and elaborating on ow te Government sould prioritise resources in transport over te next ten years. Te relative processes are led by te Ministry of Transport and Communications, and oter stakeolders involved are county councils, and road, rail, aviation, sea and municipal autorities. Te NNTP may focus on te strategic level of te decision making, toug large infrastructure projects are also included in te specific plan. In addition, te Planning and Building Act foresees tat public planning sould promote sustainable development and public participation. Te Ministry of Environment is te dominant administrative responsible for te national planning and monitoring of te Act, wile te county planning as to be documented in Regional Master Plans, wic are obligatory for all municipalities (ttp:// Counties also make county transport plans, and tey may cooperate, eiter wit oter counties on larger regional transport plans or wit city areas in te development of urban transport plans, respectively. As far as financing is concerned, te main post at te state level is te annual National Budget (ttp:// dokumenter/pdf/ gulbok.pdf). Building and maintenance of state roads are financed by te National Budget and tolls, including some cases of Public- Private Partnersips. Bot investment and maintenance of rail infrastructure are financed at a national level, over te National budget, wile aviation infrastructure and terminals are financed troug aviation taxes and commercial incomes. Te ownersip status of terminals defines financing, for example, arbours/ports are owned by municipality, bus terminals are mostly owned by county or municipality (in cities), altoug in some cases still owned by te local bus company, etc. Finland In Finland, for large projects tat ave national importance and state funding, te initiative for planning is developed by te Ministry of Transport and Communications. Te plan is ten furter developed in te Transport Agency togeter wit te regional autorities i.e. cities or oter municipalities. Focusing on regional planning, te Regional Land Use Plan defines te use of areas needed for particular purposes and te principles of urban structure in accordance to te regional development. Te next step foresees tat te planning proceeds to a more detailed Master Plan for eac region, tus te Town Plan. Te regional autorities, cities and oter municipalities forming te Regional Councils, 3

28 Transport and Telecommunication Vol. 4, No, 03 togeter wit te state autorities i.e. Centres for Economic Development, Transport and te Environment are responsible for te planning. Transport planning is a part of te general planning procedure, i.e. Land Use and Building Act 3/999, amendment /003 (ttp:// pdf); Land Use and Building Decree 895/999 (tpp:// /en pdf), and Decree on te Openness of Government Activities and on Good Practice in Information Management 030/999 ( E PDF), etc. Among te plans concerning transport is te national strategic plan Transport 030 Major Callenges, New Directions prepared by te Ministry of Transport and Communications, and included a wide-range public consultation tat was carried out in te regions. Te aim of te specific plan was to stimulate a broad-base discussion regarding te direction and framework of transport policy, in order to be used for te documentation of an action plan tat will set out te critical measures required for te next ten years and to specify te appropriate measures for te integration of te transport system (ttp:// As regards te transport financing scemes, te maintenance of public roads, railways and maritime and inland waterways is financed by central government resources, wile major transportnetwork development projects are decided by Parliament. Suc projects are granted a single contract autorization covering all te project s construction costs. New projects are included in te Government s budget under te item for transport network development. Czec Republic In 005, te Czec Government approved a new Czec Republic s Transportation Policy for tat sets strategic goals for transport and transportation networks under te scope of te full Czec membersip in te EU, including priorities suc as te improvement of transport safety, te support of regional transport development, te assurance of financing transport, te provision of ig quality transport infrastructure, etc. (ttp:// In addition, te Operational Programme Infrastructure (OP Infrastructure) for te programming period was developed in compliance wit EU regulations and works as te main document regarding te coesion of transport policy in Czec Republic, formulating seven priority axes; tus, axes and 3 for railway modernization, axes and 4 for igways and first class road construction, axis 5 concerning subway and transport operational system development in Prague, axis 6 for multimodal freigt transport and river transport development and axis 7 regarding tecnical assistance. Most of te approved projects were submitted under te 4t axis and te igest sares of funding were allocated in projects of te axes and (ttp://pernerscontacts.upce.cz/8_00/ Hajek.pdf). From te side of financing, transport infrastructure investments and sustainable transport are funded by te EU coesion policy for te period Te relevant budget for te Czec Republic is distributed among eigt sectoral operational programmes, seven regional operational programmes (ROPs), six operational programmes of te European Territorial Cooperation objective and two operational programmes designed for Prague. In tis framework of te operational programmes, an important priority was given to transport (ttp:// Greece Te current and future planning and financial processes and scemes in te transport sector in Greece are part of te National Strategic Reference Framework (NSRF) of te period Te Ministry of Economy and Finance (MEF) is responsible for elaborating te NSRF proposal for and for coordinating all relevant procedures. In te framework of NSRF, te programming documents regarding te transport sector include te Operational Programme Accessibility Improvement, te Transports Development Plan and twenty-year plan, and te National Port Policy (). Te Operational Programme Accessibility Improvement (OP-AI) includes interventions tat enance te improvement of accessibility bot regionally and internationally (i.e. trans- European interconnections). For example, in te road sector, empasis is given to te continuation of te construction of te PATHE and Egnatia motorways, te promotion of combined transport troug te implementation of te Trans-European Network for Motorways and te improvement of te regional road network and interconnections of transport networks (). On te oter and, te main objective of te Transports Development Plan (TDP) is to formulate a documented basis for te documentation of a long-term development vision, setting focused priorities suc as te integration of a national transportation system, te promotion of te country as te central node of Sout-Eastern Europe and connection link of East Mediterranean and te implementation of an improved (re)structuring of 4

29 Transport and Telecommunication Vol. 4, No, 03 te planning, control and surveillance transport system (). Lastly, te National Port Policy (NPP), formulated in 006 by te Ministry of Maritime Affairs, Islands and Fiseries, igligted te need for improving productivity, efficiency and quality of te provided services, te adjustment of te maritime system into dominant transportation trends and needs, te establisment of a friendly and flexible business environment, and te armonization and efficiency of te private maritime settlements and ports of te country (3). Regarding te financing processes in Greece, te majority of te transportation projects are cofinanced from EU, as well as from national resources (for example in cases of expropriations, etc.). In addition, private funding of projects is also foreseen. Oter funding scemes developed in specific transport investments and projects are te Public-Private Partnersips (PPPs) and joint ventures. Te funding process of transport projects is based on teir classification in tree categories. Projects of priority A are proposed for funding in te programming period , projects of priority B are also proposed for funding in te programming period , and projects of priority C are proposed for funding in te next 0 years (). 4. Planning and Financing Processes in Specific European Terminals Te findings of te questionnaire survey conducted in te specific European terminals are presented in te following paragraps, under five pillars: railways, ports, airports, bus terminals and freigt terminals. Railways Te first terminal refers to te Hellenic Railway Organization (OSE S.A.), wic is a stateowned organization, providing rail passenger and freigt services. National autorities are responsible for all stages of te planning and financing processes. Regarding te developed financing scemes, in te cases of land acquisition, engineering and construction, te funding comes from direct investment and European funds, wile management, operation/maintenance and control relies on direct investment, only. It as to be mentioned tat all companies ave an equal access to te terminal. Te second terminal regards te rail terminal Bratislava in Czec Republic, in wic te national rail manager of infrastructure (Železnice Slovenskej republiky ŽSR) owns te infrastructure, and te national rail operator (Železničná spoločnosť Slovensko a.s. ZSSK) operates te station. Te City Council and te Regional Municipality are, also, involved in te planning processes. More specifically, regarding te planning process, national, regional and local autorities and private actors, responsible for taking initiatives for investments, national, regional and local actors are involved in te stages of tecnical specifications, evaluation/selection and procurement, wile te feasibility study and construction are under te supervision of private actors. In addition, customers and demand side actors are responsible for te operations and maintenance, national and local autorities for te regulations, wile in te case of exploitation and ownersip, land is under te responsibility of national regional/local actors, infrastructure under national autorities and private actors, and operations under private actors, respectively. As far as te financing processes are concerned, national, regional and local autorities are responsible for land acquisition, wic is directly invested, wile te engineering/design, management and operation/maintenance are under te responsibility of private-private partnersips (PPPs) and are also directly invested. In construction, national and regional autorities, as well as PPPs are involved, wile funding, in tis case, is made troug direct investments, PPPs and European resources. Regarding control, tis stage is under te scope of regional and local autorities and is funded by deferred investments. Lastly, it as to be noted tat all operators ave an equal access to te terminal, wile te most important barriers tat may occur are dealt wit in te planning process. Ports Examining te planning and financing processes in Tessaloniki Port Autority S.A. (TPA S.A.) in Greece, te port works under a concession agreement between te state and te TPA S.A, according to wic TPA S.A. as te exclusive rigt to use and exploit land, buildings and facilities of te port, owed by te state. All stages of te planning process are under te responsibility of te TPA S.A., expect te regulatory framework, wic is under te scope of te state. Regarding te financing process at te port, all stages are under te responsibility of TPA S.A., and te funding sceme is exclusively direct investment. Also, te public participation is foreseen troug te submission of proposals to te Ministry of Mercantile Marine and Aegean Sea, te announcements to TPA S.A. website for public consultation, occasionally consultation wit te Municipality, wile international and national public tenders for procurement are also foreseen, in accordance wit te European legislation. In addition, it as to be 5

30 Transport and Telecommunication Vol. 4, No, 03 mentioned tat, in te case of big infrastructure projects, te ministry of Mercantile Marine and Aegean Sea, as to coordinate wit te procedures for te issuance of Common Ministerial Decisions, and in case of international calls for tender wit budget over euros, a ruling Court of Auditors is requested before te signing of te contract. All decisions are made by te TPA S.A. board of Directors. Finally, te most important barriers, bot in planning and financing processes, refer to finance and legislation. In te case of te Norwegian Coastal Administration, wic is te national agency for coastal management, maritime safety and communication in Norway, te initiative for investments, procurement and feasibility study are under te responsibility of national and regional/local autorities, te tecnical specifications and te regulatory framework under national autorities, te construction under private actors, te evaluation/selection under national, regional and local autorities and private actors, and te operation/maintenance under regional/local and private actors. In te stage of exploitation/ownersip, regional and local autorities are responsible for land, and national, regional and local autorities for operations. In terms of te financial process, te state is involved in all stages, wile regional and local autorities are responsible for engineering/design, construction, management, operation/maintenance and control. Customers and actors from te demand side are also involved in te stages of construction and operation/maintenance. Also, te financing sceme tat is developed in all stages of te financial process is direct investments, and PPPs in te cases of engineering/design, construction and operation/maintenance. Lastly, it as to be noted tat public participation is formally ensured troug earings and meetings wit stakeolders in te early stages of te planning, wile important barriers are indicated in te coordination and initial stages of te planning process. Airports Investigating te planning and financing processes from te side of airports, te first terminal regards te Oslo Airport Gardermoen, in Norway, in wic, te Oslo Airport LTD and te moter company Avinor LTD, are responsible for all stages of planning, excluding operation/maintenance and regulatory framework, wic are under te autority of te Oslo Airport. As far as te financing process is concerned, Avinor LTD is responsible for te funding of land acquisition, engineering/design, construction and management, wile te Oslo Airport is involved in engineering/design, construction, management, operation/maintenance and control. Te land acquisition is funded troug state and private loans, wile te resources for engineering/design, construction and management are provided by te state and private loans and direct finance from te Oslo Airport. Oslo Airport also finances operations and maintenance. Te second air transport terminal refers to te Prague Airport Ruzyn, wic is a joint stock company. Te exclusive responsibility of initiative for investments, tecnical specifications, procurement and evaluation/selection is under national autorities, wile te feasibility studies, construction and operation/maintenance is under private actors, respectively. Te regulatory framework is under te scope of national and regional/local bodies. In terms of exploitation/ownersip, te part of land in under te supervision of national autorities, and te infrastructure and operations are under national, regional and local actors. Regarding te financing process, national autorities are involved in land acquisition, construction, management and control, and PPPs in all stages. Specifically focusing on land acquisition, engineering and management, te financing sceme is direct investment, in construction and operation/ maintenance te sceme is direct investment and PPPs, wile control is funded by deferred investments. Bus terminals From te scope of public transport, in te Urban Pubic Transport Organization of Tessaloniki, Greece, te national autorities are involved in te tecnical specifications and te infrastructure part of te ownersip, and te regional/local autorities in te regulatory framework, land and infrastructure. Also, private actors are responsible for all stages of te planning process, except of te regulatory framework and te construction in wic manufactures are involved. As far as financing concerned, regional and local autorities are involved in land acquisition, engineering/design and construction, and PPPs in all stages. Te funding of land acquisition is under direct and indirect investments and PPPs, wile all oter stages are funded via direct investments. Te most significant barriers indicated in te specific organization regard coordination and legislation in te planning process, and resources finding in te financing planning process. In Czec Republic and te Brno bus station, two categories of actors involved in te planning process are identified: regional/local autorities and private actors. Te first category is indicated in te initiative for investments, tecnical specifications, procurement, feasibility study, evaluation/selection, 6

31 Transport and Telecommunication Vol. 4, No, 03 regulatory framework and land in exploitation/ownersip. On te oter and, private actors are involved in initiative for investments, procurement, feasibility study, construction, operation/maintenance, as well as in te sectors of infrastructure and operations in te stage of exploitation/ownersip. In terms of financing and te relative allocation of responsibilities among te involved actors in te specific terminal, national and regional autorities are indicated in land acquisition and engineering/design, and PPPs in te rest stages, tus, construction, management, operation/maintenance and control. Te funding sceme in land acquisition is direct investment, in construction European funds and PPPs, wile te rest stages of te process are financed via PPPs. Lastly, it sould be noted tat te most important barriers are indicated in coordination and control in te planning process, and in resources finding and control in te financing planning process, respectively. Freigt terminals Te last terminal investigated in te present paper regards te private freigt terminal SINDOS S.A., in Greece. Due to te private status of te terminal, private actors are mostly involved in all stages of te planning process, excluding te regulatory framework and te evaluation/selection, for wic national autorities are responsible. National autorities are also met in te state of te stage of initiative for investments and procurement. Regarding te financing process, te responsible actors are eiter national autorities, i.e. land acquisition, engineering/design and construction, or PPPs in all stages. Direct investments are developed in all stages of te financing process, and European funds, are, also, used for land acquisition, engineering/design and construction. It as to be mentioned tat during te planning process, procurement and terms of reference ad been publised widely and tenders were invited. Also, te selection of te property on wic te terminal was developed was based on te existing land uses, establised by te national regulations (i.e. industrial zone), and te specific site was selected so as to ave direct access to te national igway network and te main railway network, wic ad to be expanded by te developers. Finally, te most important barriers are indicated in legislation. Summary and Discussion Te aim of te present paper was to investigate te planning and financing scemes developed in long-sort distance interconnection in representative European countries, Spain, France, Norway, Finland, Czec Republic and Greece, as well as in specific European terminals, and to assess ow tese processes are linked to te decision-making processes at a local, regional, national and European level. For te assessment of te commonalities and differences in te above processes in te case of te European countries, a template was developed including key points addressing planning and financial processes in eac country, wile, in te case of terminals, a questionnaire survey was conducted, addressing issues suc as te identification of te involved stakeolders, public participation, funding scemes, indication of barriers and suggestion of metods for teir alleviation, etc. Analysing and assessing te data collected for te planning and financial processes in te representative countries, it was indicated tat te strategic sceme of te planning processes followed in te long and sort-distance interfaces is usually a National Transportation Plan NTP, wit a time orizon usually 0 30 years. Te Ministries of Transport are te main responsible bodies for te preparation and coordination of te NTPs. In case tat te NTP is part of an Operational Programme (OP) (as in Greece for example) ten, apart from te ministry, te relative managing autorities of te OP play also a significant role in te preparation and management of te NTP. As far public participation is concerned, it is common an open procedure (on-line call for proposals, etc.) to be followed before te finalization of te NTPs. Investigating te planning scemes at a regional level, common processes were also observed, e.g., te preparation of Regional Transport Plans (RTPs), wic are usually parts of te NTPs, toug focusing on regional needs. At an urban or local level, most of te large cities in European countries, prepare Urban Mobility Plans (UMPs), focusing on providing daily services to users (i.e. comfort, safety, etc.). Regarding to te planning scemes identified at specific terminals, tese vary from country to country and from mode to mode. In general, te status ownersip (private, stated-owned, public-private, etc.) sets te parameters bot for te planning and financial processes, wile EU and national legislation also affects te preparation and implementation of te plans at passenger and freigt terminals. Te planning time orizon is usually 0 years, and te content of te plans is strongly dependent on te location (topograpy and tecnical appropriateness, complementary activities in te surrounding areas, etc,), as well as of te accessibility of te terminal (distance from ports, airports, railways, urban and commercial centres, nearest igways, etc.). 7

32 Transport and Telecommunication Vol. 4, No, 03 Finally, te review of te financial processes followed in European countries, sowed tat majority of transportation projects are co-financed by EU, as well as by national resources and private funding (example.g., in cases of expropriations, etc.). Te relative plans, i.e. NTPs, RTPs or UMPs, may define te allocation of co-funding of te projects. Te main responsible for te financial management is te responsible ministry of finance and economy in eac country, wic as te general supervision and sets te priorities in funding several projects. A rater expanded funding sceme in EU countries is Public- Private Partnersips (PPPs), tus, types of cooperation between te public and private sector based on a contract between te two bodies. Te most significant problems tat were met in te above countries and terminals are related to te involvement of more tan one body in te planning and financing stages of te decision-making process in te interconnection of long-sort distance transport, and te relative conflict of interests of te involved stakeolders. Potential solutions for te alleviation of tese problems include te establisment of a well documented cooperation among te involved stakeolders, te distinction of ownersip status from operation for te insurance of te equal access to all, and te pursuance of Public-Private Partnersip models for te confrontation of complex local and regional problems and financing issues. Acknowledgements Te present paper is based on te researc tat as been conducted till now in te framework of te CLOSER project (Connecting Long and Sort-distance networks for Efficient transport), wic is co-funded by te European Commission witin te 7 t Framework Programme. Te autors would like to tank bot te consortium of te project and te European Commission. References. European Commission. (00). European Transport Policy for 00: time to decide. Wite Paper. CEC.. Commission of te European Communities. (00). COM (00) 307 final. European Transport Policy for 00: time to decide. European Communities, Brussels. 3. European Commission. (006). Keep Europe Moving. Sustainable mobility for our continent. Midterm review of te European Commission s 00 transport. Wite Paper. ISBN Luxemburg: Office for Official Publications of te European Communities. 4. European Commission. (0). COM (0) 44 final. Roadmap to a Single European Transport Arena towards a competitive and resource efficient transport system. Wite Paper. European Commission. 5. Natanail, E., Adamos, G., Parra L., Ruiz-Ayucar, E., L Hostis, A., Blanquart, C., Olsen, S., Cristiansen, P., Osland, O., Järvi, T., Svedova, Z. & Zan, B., (0). Analysis of te Decision- Making Framework. In E. Natanail & G. Adamos (Eds.) Deliverable D4.. CLOSER Connecting LOng and Sort-distance networks for Efficient transport. CLOSER Project. Brussels, Belgium: CLOSER Consortium. 6. Adamos, G., Natanail, E. & Zacaraki, E. (0). Developing a Decision-Making Framework for Collaborative Practices in Long-Sort Distance Transport Interconnection. Procedia Social and Beavioral Sciences, Volume 48, 0, pp , Elsevier. 7. May, A. D., Karlstrom, A., Marler, N., Mattews, B., Minken, H., Monzon, A., Page, M., Pfaffenbicler, P. C., Seperd, S. (003). Deliverable N o 5. Decision Maker s Guidebook. PROSPECTS Project. Brussels, Belgium: Prospects Consortium. 8. Emberger, G., Pfaffenbicler, P., Jaensirisak, S. & Timms, P. (009). Ideal decision-making processes for transport planning: A comparison between Europe and Sout East Asia. Transport Policy, 5, Ward, D. (June 00). Stakeolder involvement in transport planning: participation and power. Impact Assessment and Project Appraisal, 9(), Szyliowicz, J. (003). Decision-making, intermodal transportation, and sustainable mobility: towards a new paradigm. Oxford, UK: Blackwell Publising, Ltd.. National Strategic Reference Framework (NSRF) , Hellenic Republic, Ministry of Economy and Finance, General Secretariat for Investments and Development, Atens, Greece, Transports Development Plan (TDP) , and twenty-year plan, Hellenic Republic, Ministry of Infrastructure, Transport and Networks, National Port Policy (NPP) (006), Hellenic Ministry, Ministry of Maritime Affairs, Islands and Fiseries, General Secretariat of Ports and Port Policy,

33 Transport and Telecommunication Vol. 4, No, 03 Transport and Telecommunication, 03, Volume 4, No, 9 38 Transport and Telecommunication Institute, Lomonosova, Riga, LV-09, Latvia DOI 0.478/ttj DISTINGUISHING BETWEEN SPATIAL HETEROGENEITY AND INEFFICIENCY: SPATIAL STOCHASTIC FRONTIER ANALYSIS OF EUROPEAN AIRPORTS Dmitry Pavlyuk Transport and Telecommunication Institute Lomonosova, Riga, LV-09, Latvia P.: (+37) Dmitry.Pavlyuk@tsi.lv A problem of distinguising between frontier eterogeneity and inefficiency is widely acknowledged in bencmarking. A special type of eterogeneity, based on te spatial structure, can significantly affect performance estimates in te airport industry. In tis researc we presented a general specification of te spatial stocastic frontier model, wic includes spatial lags, spatial autoregressive disturbances and spatial autoregressive inefficiencies. Maximum likeliood estimator as been derived for tis model. Applying te suggested model specification to te European airports dataset, we discovered presence of significant spatial eterogeneity, wic leads to biased estimates of efficiency, received using classical models. Keywords: airport efficiency, spatial stocastic frontier, maximum likeliood estimator Introduction Airport bencmarking attracted significant scientific attention after industry liberalisation in te nineties [], []. Tere are more tan a undred researc papers, publised during last two decades, and devoted to airports efficiency estimation. Te most significant reports are te Global Airport Performance Bencmarking Reports produced by Air Transport Researc Society [3], te Airport Performance Indicators and Review of Airport Carges reports by Jacobs Consulting, te Airport Service Quality programme by Airports Council International. Some local autorities, wic control te airport sector, also provide teir own bencmarking reports, e.g. Civil Aviation Autority (UK) [4], and oters. Many related researces are also executed witin te bounds of te German Airport Performance researc project, a joint study between tree German universities. Despite intensified researc of efficiency in te airport industry and obvious locational issues, spatial effects are rarely included into consideration. Spatial interactions between European airports and spatial eterogeneity of te industry structure are widely acknowledged [] and sould be included into airport bencmarking tecniques [5 7]. Spatial relationsip is usually presented in a form of spatial competition. Teory of spatial competition is well developed, but rarely applied to airports. Taking spatial interactions between airports into models is critically important for estimation of airports efficiency levels. A wide range of instruments like overlapping catcment areas, a network connectivity index and oter were suggested, but te spatial econometric models can be igligted as te most teoretically supported metodology. Spatial eterogeneity is also a very important drawback in airport efficiency researc. Regionspecific settings can significantly affect airports activity, but teir inclusion into a model is not straigtforward. Tere are some sources of regional eterogeneity of airport activity: - Climate. Airport activity can be significantly affected by a climate. For example, snow-belt airports ave to spend additional efforts on runway service, wic reduce teir production in relation wit airports located in a region wit softer weater conditions. Mapping tis difference out te model will lead to underestimated values of snow-belt airports efficiency. - Economics. Economic situation in European countries is very eterogeneous. Significantly different income per capita and price levels define different demand to air transport. - Region attractiveness. Regions also are not equal in relation to teir demand for air transport. Business activities, required air fligts, tourism attractiveness significantly vary across Europe. A standard approac to include tese factors into te model is based on a set of region-specific dummy variables, and looks weak in case of a complex spatial structure. Spatial effects are usually not limited wit country borders, so using of an administrative division in tis case is not well-grounded. Neverteless, spatial structure sould be included into efficiency estimation to prevent a bias in efficiency estimates. 9

34 Transport and Telecommunication Vol. 4, No, 03 Te most commonly used parametric approac to estimate efficiency levels is based on te stocastic frontier [8]. Classical specification of te stocastic frontier represents an optimal ratio of a set of used resources to produced outputs. Tis approac considers all deviations from te frontier as unit inefficiency and doesn t take possible eterogeneity into account. Tere are a number of different metods proposed to distinguis eterogeneity and inefficiency witin te stocastic frontier models. Te most frequently used approac is based on inclusion of variables, describing eterogeneity into te model (observed eterogeneity). For airports it can be geograpical regions, ownersip [9], [0], government regulation [], and oters. Anoter possible way to distinguis eterogeneity and performance relies on te time factor and assume tat eterogeneity is more stable over time tan inefficiency []. Despite a significant number of empirical applications, given metods ave a number of sortcomings, generally related wit teir insufficient flexibility and requirements for initial assumptions. Spatial proximity is rarely used in suc models, wic generally leads to lose of information in case te spatial structure plays a role. Development of spatial econometrics [3] allows including spatial eterogeneity and spatial relationsips into parametric models in a undisguised and flexible way. However spatial specification of te stocastic frontier model is insufficiently researced. In tis researc we consider a general form of te spatial stocastic frontier model wit all types of spatial components spatial lags, spatial autoregressive disturbances and spatial autoregressive inefficiencies. A set of popular econometric tecniques (maximum likeliood, two-step least squares, general metod of moments) can be adapted estimation of tis model. We develop a maximum likeliood estimator for different forms of te stocastic frontier model. Applying te developed estimator to a data set of European airports, we analysed an influence of spatial components on estimated airport efficiency. Specifications of a Spatial Stocastic Frontier Model A classical stocastic frontier model is usually presented in a matrix form as [8]: y = Xβ + ε, ε = v u, u 0, () were y is an (n x ) vector of a dependent variable, output (n is a size of te sample); X is an (n x k + ) matrix of explanatory variables, inputs (k is a number of explanatory variables); β is a ( x k + ) vector of unknown coefficients (model parameters); ε is an (n x ) vector of composite error terms; v is an (n x ) vector of independent identically distributed (i.i.d.) error terms; u is an (n x ) vector of inefficiency terms wit non-negative values. Te classical stocastic frontier model doesn t include any spatial dependencies and assumes tat all objects in a sample are independent. Tis assumption is too restrictive in some practical cases. Spatial effects can be presented almost in all components of te classical model: - spatial influence of neigbours output values on a given unit s output (spatial lags); - spatial influence of neigbours input values on a given unit s output; - spatial relationsip between neigbour unit s error terms (spatial eterogeneity); - spatial correlation between efficiency of neigbour units. We define a general spatial stocastic frontier model, including all tese effects into te classical stocastic frontier model specification: y = ρw y + Xβ + γw X + ε, ε = v u, v = ξw ~ 3v + v, u = ηw u + u ~, u 0, 4 () 30

35 Transport and Telecommunication Vol. 4, No, 03 were W y is a spatial lag vector of output values (wit a coefficient ρ); W X is a spatial inputs-output lag vector (wit a coefficient γ); W 3 v are spatial errors (wit a coefficient ξ); W 4 u are spatial inefficiency lags (wit a coefficient η). Matrices W, W, W 3, and W 4 represent levels of spatial dependency between units (spatial weigts) and can be different for every spatial component. All spatial weigts matrices ave zero values on te main diagonal (to prevent self-dependency). Construction of tese matrices is usually researc-specific and can be based on geograpical distances, travel times, etc. Estimation of te general spatial stocastic frontier model s parameters is a complicated task, wic is related wit identification problems, computation performance issues and requires a significant volume of data. In tis researc we consider two special cases of te general spatial stocastic frontier model. Let s apply te following constraints on te general spatial stocastic frontier model: γ = 0, (3) ξ = 0, η = 0. (4) (5) Under tese constraints te model includes only spatial lags for unit s outputs, all oter spatial effects are excluded: y = ρw y + Xβ + ε, ε = v u, v = v~, u = u ~, u 0. (6) Following LeSage [4] notation for naming of spatial models, tis model is a mixed first-order spatial autoregressive-regressive stocastic frontier model. We will refer tis model as a spatial autoregressive stocastic frontier (SARSF) model. Anoter model we consider in tis researc includes spatial relationsip of a symmetric error term v as well as spatial lags. Suppressing te constraint (4), we obtain te following model specification: y = ρw y + Xβ + ε, ε = v u, v = ξw ~, ~ v + v u = u, u 0. (7) We will refer tis model as a mixed first-order spatial autoregressive-regressive stocastic frontier model wit spatial autoregressive disturbances (SARARSF). Maximum Likeliood Estimator for Spatial Stocastic Frontier Models A wide range of statistical metods is used for estimation of spatial model parameters. Te most popular are maximum likeliood estimator [3], [4], two-step least squares [5], and generalised metod of moments [6], [7]. In tis researc we derive maximum likeliood estimators for SARSF and SARARSF models. Te maximum likeliood estimator requires an assumption about distributions of error and inefficiency components. Te distribution of te symmetric error term v is usually set to normal, and te distribution of te non-negative inefficiency term u is selected from alf-normal [9], truncated normal [0], or gamma []. We consider te simplest normal-alf-normal type of te composite error term ε: Derived estimators are presented and discussed on te t International Conference Reliability and Statistics in Transportation and Communication (RelStat') [8] 3

36 Transport and Telecommunication Vol. 4, No, 03 ~ v = v ~ N(0, σ v~ I), ~ + u = u ~ N (0, σ ~ I). u (8) Te probability density function for tis case is well known: ε ελ f ( ε ) = ϕ Φ, (9) σ σ σ were σ u~ σ = σ u~ + σ v~, λ =, σ ~ v φ and Φ are standard normal probability density and cumulative distribution functions accordingly. Maximum Likeliood Estimator for te SARSF Model Derivation of te maximum likeliood estimator formula for te SARSF model (6) is quite straigtforward. According to te model specification, te composite error terms vector can be expressed as: y = ρw y + Xβ + ε ε = ε = ( I ρw ) y Xβ. y ρw y Xβ Using te multivariate cange of variables formula and te Jacobian matrix, we can produce te probability density function for y: f f ( y) = f ( ε ) ε = y ε ε ελ = Φ y. (0) σ σ σ ( y) f ( ε ) det = ϕ det( I ρw ) LogL Now te log-likeliood function can be easily obtained: T n ε ε ελ = det σ i= σ ( ρ, β, σ, y, X, W ) C nln( σ ) + ln Φ + ln( ( I ρw )) Maximum Likeliood Estimator for te SARARSF Model. () We follow te procedure, described in Kumbakar and Lovell [8] to produce te probability density function and likeliood function for te SARARSF model (7). Initial model specification includes te distribution of te error term in an implicit form: v = ξ W ~ v + v, ~ v ~ N(0, ~ I). v = (( I ρw ) y + Xβ ) σ v Straigtforward transformations give us: v ξw v = v ~ v ξw v = ( I ξw ) v ~ ε det y y = ~ v ~ ( I ρw ) ( I ξw ) v ~ v ~ T ( I ξw ) ) ( I ξw ) v ~ N(0, Σ),were Σ = σ Σ = σ. () 3

37 Transport and Telecommunication Vol. 4, No, So te error term as a multivariate normal distribution wit a covariance matrix Σ and its respective probability density function is given as: () ( ) ( ) Σ Σ = v v v f T n exp det π. (3) Te alf-normal probability density function is given as: ( ) < = 0 0, 0, exp ~ ~ u u u u u f u T n u σ π σ. (4) Assuming u and v components are independent, te joint normal-alf-normal probability density function is a product of functions (3) and (4): ( ) ( ) ( ) ( ) ( ) ( ) ( ). exp det exp exp det, ~ ~ ~ ~ Σ + Σ = = Σ Σ = = v v u u u u v v v f u f v u f T u T n u u T n u T n σ πσ σ π σ π Since ε = v u: ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ). exp det exp det, ~ ~ ~ ~ Σ + Σ + Σ + + Σ Σ = = + Σ + + Σ = ε ε ε ε σ πσ ε ε σ πσ ε T T T u T n u T u T n u u u u I u u u u u u f Straigtforward transformations give a simplified form of te joint probability density function: ( ) ( ) ( ) ( ) ( ) ( ), exp exp det, ~ Θ Ω Σ = ε ε με με πσ ε T T n u u u u f (5) were.,, ~ ~ = ΩΣ ΣΘ Ω = + Σ Θ = μ σ σ u u I Te marginal density of ε is obtained by integrating u out of f(u, ε): ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ), exp exp det det det exp exp det, 0 ~ 0 0 Θ Φ Ω = = Θ Ω Ω Ω Σ = = Θ Ω Σ = = ε ϕ με ε ε π με με π σ ε ε με με πσ ε ε n T n T n n u T T n u du u u du u u du u f f

38 Transport and Telecommunication Vol. 4, No, 03 were φ(x, mean, covariance) and Φ(x, mean, covariance) are multivariate probability density and cumulative distribution functions wit a mean vector mean and a covariance matrix covariance. So finally te probability density function of te composite error term ε is obtained: n ( ε ) = ( Φ( 0, με, Ω) ) ϕ( ε, 0, Θ) f. (6) Tis finding matces te results presented in []. Using te density function (6) and following te same logic as in (0), we obtained te respective log-likeliood function for te SARARSF model: LogL( y β, σ ~, σ ~, ρ) = C + ln( Φ( 0, με, Ω) ) + ln( ϕ( ε, 0, Θ) ) + ln( det( I ρw )), (7) were ~ Σ = σ v ~ Σ, ~ Θ = σ u ~ I + σ ~ v Σ, ~ Ω = σ u ~ σ v ~ ΣΘ, ~ μ = σ ~ ΩΣ. v u v Maximisation of te log-likeliood function wit respect to its parameters is a separate computational problem. Application to Airports We applied bot SARSF and SARARSF specifications of te spatial stocastic frontier model to a data set of European airports. Tere are many different approaces to understanding of airport business, a set of used resources and an output of airport activity [3]. In tis researc we used a number of transferred passengers of a main result of airport activity and infrastructure units (gateways, ceck-ins) as airport resources. Te production function used in tis researc is estimated in te form (SARARSF model): ( Passengers) = β + ρw ln( Passengers) + β ln( Runways) + β ln( CeckIns) ln 0 ~, ~ v = ξwv + v v ~ N(0, σ v~ I) + u ~ N (0, σ ~ I), u 3 + v u, were Passengers is a total number of passengers carried by an airport (bot departure and arrival); Runways and CeckIns are numbers of airport s runways and ceck-ins respectively. Oter airport infrastructure caracteristics are excluded due to a multicollinearity problem. A matrix of spatial weigts W was constructed on te base of Euclidean distances. Realising all sortcomings of tis approac, we tink tat general influence of spatial effects will be estimated correctly. Data Te study data set includes caracteristics of European airports in 009. Te caracteristics include: - A number of passengers carried (direct transit passengers are excluded). Tis indicator is used as te main output of airport s activity. - Airport infrastructure a number of ceck-in facilities, gates, runways, and parking spaces are used as input resources of airports activity. Te full dataset is collected from te Eurostat database [4]. Descriptive statistics of te collected dataset is presented in te Table. (8) 34

39 Transport and Telecommunication Vol. 4, No, 03 Table. Data descriptive statistics Total number of passengers carried, tousands Number of runways Number of ceck-ins Min Median Mean Max Estimation results Tree different model specifications were tested for te data set: - classical stocastic frontier (SF estimates); - spatial autoregressive stocastic frontier (SARSF estimates); - spatial autoregressive stocastic frontier wit spatial autoregressive disturbances (SARARSF estimates). Te results are presented in te Table 3. Table. Estimation results of different model specifications * SF SARSF SARARSF β (< E-6) (< E-6) (< E-6) β (7.3E-07) (9.4E-07) (5.7E-07) β (< E-6) (< E-6) (< E-6) σ v σ u ρ (9.4E-0) (9.0E-0) ξ (4.7E-08) * Values in brackets are significance Standard SF estimates sow significant inefficiency in data (inefficiency variance σ u = 0.48, wic is comparable wit te error term variance σ u ). Moran s I is a popular test statistics for spatial dependence [5]: T e We MoranI = ~ i. i. d. normal T. (9) e e Applying te Moran s I test to estimated efficiency levels of te standard SF model, we strongly reject a ypotesis about absence of spatial relationsips in data (see Table 3). Note tat observed Moran s I value is positive, wic indicates positive spatial relationsip. Table 3. Moran s test for spatial dependency in estimated efficiency values Null ypotesis Alternative ypotesis Moran s I value p-value (two-sided test) No spatial dependency Presence of spatial dependency Null ypotesis is rejected 35

40 Transport and Telecommunication Vol. 4, No, 03 Spatial distribution of SF estimates of efficiency levels is presented on te Figure. We can note tat efficiency levels are not distributed uniformly, aving areas wit iger (darkk circles) and lower (ligt circles) values. Figure. Spatial distribution of European airports efficiency Te SARSF model sows insignificant spatial lags and significant inefficiency. Taking spatial eterogeneity into account (SARARSF model) we produce alternativee results. Tee SARARSF model demonstrates significant spatial eterogeneity in data (wic is expected for te airport industry), and tis spatial component supplants inefficiency it becomes insignificant. Tis result is expected due to incompleteness of te production function components ( only infrastructure metricss are included), but from our opinion te general findings are very important for furter analysis. Generally, imperfection of te production frontier can produce incorrect conclusionss about te inefficiency term, wen including spatial components into te model can partly improve te situation. Estimates of input elasticises (coefficient β for a number of runways andd coefficient β for a number of ceck-ins) are quite stable for all models, wic is reasonable in absence of significant spatial lags. Conclusions In tis researc we presented a generall specification of te spatiall stocastic frontier model, wic includes spatial lags, spatial autoregressive disturbances and spatial autoregressive inefficiencies. Tese spatial components describe present spatial relationsips and spatial structure of studyy units and critically important for econometric models, specifically for efficiency estimation. Maximum likeliood estimator for te spatial autoregressive stocastic frontier wit spatial autoregressive disturbances is derived. We applied te suggested models wit spatial components to a data set of European airports. We discovered significant positive spatial relationsip in airport efficiency levels, estimated using te standard stocasticc frontier model (witout spatial components). Applying te stocastic frontier model wit spatial lags only, we also received significant inefficiency in data. However application of te most general model (wit bot spatial lags and spatial eterogeneity) led us to te opposite conclusion 36

41 Transport and Telecommunication Vol. 4, No, 03 we discovered significant spatial eterogeneity, wic was considered as inefficiency in te previous models. We regard tese results as an important evidence of necessity of spatial components in stocastic frontier models. Non-inclusion of spatial components into bencmarking models can lead to significant biases of frontier parameters and efficiency levels estimates. Acknowledgement Te autor is grateful to Jiaxian Lin, Sout Cina University of Tecnology, for providing materials and very useful comments on te maximum likeliood estimation of te spatial autoregressive stocastic frontier model. References. Adler, N., Liebert, V. (0). Competition and regulation (wen lacking te former) outrank ownersip form in generating airport efficiency. Presented at te GAP worksop Bencmarking of Airports, (p. 7). Berlin, Germany.. Pavlyuk, D. (0). Airport Bencmarking and Spatial Competition: A Critical Review. Transport and Telecommunication, 3(), Oum, T. H., Yu, C., Coo, Y. (0). ATRS Global Airport Performance Bencmarking Project. Te Air Transport Researc Society. 4. Te Use of Bencmarking in te Airport Reviews. (000). London, UK: Civil Aviation Autority. 5. Barros, C. P., Managi, S., Yosida, Y. (008). Tecnical Efficiency, Regulation, and Heterogeneity in Japanese Airports, Working Paper 43/008/DE/UECE. Lisbon, Portugal: Scool of Economics and Management. 6. Barros, C. P. (0). Performance, eterogeneity and managerial efficiency of African airports: te Nigerian Case. Lisbon, Portugal: Scool of Economics and Management. 7. Liebert, V. (00). External eterogeneity in airport bencmarking and efficiency analysis. Bremen, Germany: Jacobs University. 8. Kumbakar, S. C., Lovell, C. A. K. (003). Stocastic frontier analysis. Cambridge: Cambridge Univ. Press. 9. Adler, N., Liebert, V. (0). Joint Impact of Competition, Ownersip Form and Economic Regulation on Airport Performance. Bremen: Jacobs University. 0. Barros, C.P., Marques, R.C. (008). Performance of European Airports: Regulation, Ownersip and Managerial Efficiency, Working Paper 5/008/DE/UECE. Lisbon, Portugal: Scool of Economics and Management,. Maligetti, P., Martini, G., Paleari, S., Redondi, R. (007). Efficiency of Italian airports management: te implications for regulation. Working Paper. Italy.. Greene, W. (003). Distinguising between Heterogeneity and Inefficiency: Stocastic Frontier Analysis of te World Healt Organization s Panel Data on National Healt Care Systems. New York: Stern Scool of Business. 3. Anselin, L. (988). Spatial econometrics: metods and models. Dordrect: Kluwer Academic Publising. 4. LeSage, J. P. (999). Spatial econometrics. Morgantown, USA: Regional Researc Institute, West Virginia University. 5. Kelejian, H. H., Pruca, I. R. (998). A generalized spatial two-stage least squares procedure for estimating a spatial autoregressive model wit autoregressive disturbances, Te Journal of Real Estate Finance and Economics, 7(), Druska, V., Horrace, W. C. (004). Generalized moments estimation for spatial panel data: Indonesian rice farming, American Journal of Agricultural Economics, 86(), Lee, L. (007). GMM and SLS estimation of mixed regressive, spatial autoregressive models, Journal of Econometrics, 37(), Pavlyuk, D. (0). Maximum Likeliood Estimator for Spatial Stocastic Frontier Models. In Proceedings of te t International Conference Reliability and Statistics in Transportation and Communication (RelStat ), Riga, Latvia (pp. 9). Riga: TTI. 37

42 Transport and Telecommunication Vol. 4, No, Aigner, D., Lovell, C., Scmidt, P. (977). Formulation and estimation of stocastic frontier production function models. Journal of econometrics, 6(), Stevenson, R. (980). Likeliood Functions for Generalized Stocastic Frontier Functions. Journal of Econometrics, 3, Greene, W. H. (990). A gamma-distributed stocastic frontier model. Journal of econometrics, 46( ), Lin, J., Long, Z., Lin, K. (00). Spatial Panel Stocastic Frontier Model and Tecnical Efficiency Estimation. Journal of Business Economics, 3(5), Doganis, R. (99). Te airport business. London: Routledge. 4. European Statistics Database. (0). Statistical Office of te European Communities (Eurostat). 5. Moran, P. A. P. (950). Notes on Continuous Stocastic Penomena. Biometrika, 37( ),

43 Transport and Telecommunication Vol. 4, No, 03 Transport and Telecommunication, 03, Volume 4, No, Transport and Telecommunication Institute, Lomonosova, Riga, LV-09, Latvia DOI 0.478/ttj MOBILE AUTONOMOUS ROBOTS A NEW TYPE OF CITY PUBLIC TRANSPORT Vasili Sut, Valery Kasyanik Brest State Tecnical University, Brest, Belarus,st. Moskovskaja, 67 P.: lucking@mail.ru Brest State Tecnical University, Brest, Belarus, st. Moskovskaja 67 P.: vvkasyanik@bstu.by In te article te problem of modern cities transport jams is taken up. We propose several variants of solution and one of te main te use of autonomous mobile robots as public transport. Tis is caused by active development of robotics. Te article contains te review of te latest acievements of tis perspective spere and tecnical caracteristic of two mobile robots for providing effectiveness of traffic. Functioning of robots takes place in confined environment wit distinct infrastructure. Keywords: robot, intellectual transport system, jam, city public transport, passenger traffic. Introduction In system of road transport in wic about /3 all volumes of transport service are carried out, in tis or tat form works up to 8 0% of able-bodied people. In te subsystem of traffic system of road transport, eac person, on te average, daily is about one our, i.e. to 7% of te active time. Tus, road transport and traffic control entering into it are one of te most important life support systems of modern society. Development of transport infrastructure, including street road network development, development of new means of transport and management of it is an actual problem of modern society. In tese conditions automated control systems for movement of transport and foot streams in te cities constantly develop. Te tendency of automatic control traffic systems can be considered in two directions. Te first direction improvement of tecnical providing control systems by transport at te expense of development: electronic computing means, equipments of data transmission, widespread introduction of radio communication in a link centre peripery including cannels of cellular communication, introduction of more perfect means of information display, expansion of transport detectors, building of intellectual opportunities and road controllers. Te second direction strengtening of carried-out functions by te transport system providing improvement of service quality for passengers, and also development of new types of public transport. Growt of load on te main igways of a street road network (SRN) of te cities leads to steady decrease of te speed of movement of transport streams and to formation of traffic block. Under traffic conditions close to a traffic block, te turn of veicles cannot go in on a stage between te next intersections tat lead to violation of an operating mode of te traffic ligt system of te previous intersection. Suc situation extending on some intersections, received te name of a network jam or network [ ] saturation. Te network jam against spasmodic growt of park of cars even more often comes in SRN of te large cities. Besides automobile transport (AT) te second consumer of an SRN resource is te city public transport (CPT) wic sare in passenger traffic is 60% (earlier tis figure was 90%). It sould be noted tat CPT uses te SRN resource more effectively. On te average in te bus, a trolleybus, a tram 50 people, wile in te car only,5 persons [3] go. To avoid in te future te collapse of transport system and to support its functioning at optimum possible level it is necessary to carry out cardinal reorganization of all transport complex including SRN, management of transport streams, development of new types of CPT. If te first position witin istorically developed city building ardly gives in to reorganization, te second and te tird are quite possible. So use of adaptive management by transport streams in SRN allows increasing capacity of a network [4 5] on 0 30%. Furter it will be necessary to limit unreasonable growt of transport streams from cars to a way of input of quotas of a trip. Everyting tat is above tis quota sould be paid. In economy it is known tat free of carge provided resource injuriously is taken away. Te sare of CPT in transport streams plying in SRN as a result will increase at te general reduction of number of transport units. 39

44 Transport and Telecommunication Vol. 4, No, 03 Given above a measure are necessary, but are insufficient. Essential expansion of capacity SRN excluding collapse, is possible on a way of essential revision of principles of management of transport streams on te basis of modern acievements in te field of informatics, telecommunication systems, collective beaviour of automobile, a robotics. All in large quantities and in a variety of qualities te robots become our reality: in art, life, production, etc. Tere are robot violinists [6], robot cooks [7], robot nannies, robot engineers, robot firemen [8], security guards and oters. Robots are indispensable in many industries. For example, te robot welders are commonly used in automobile production. Tere are robots engaged in painting. In te electronics industry robots are used for soldering microscopic wires, placement of integrated circuits on circuit boards in monitoring and diagnosis of completed devices, and more. Robots are used in traffic management of veicles [9]. Tese specialized robots make te same ig-precision work every day. For a uman, suc work is boring and tedious te monotony leads to fatigue, wic causes errors. Production errors are decreasing te labour productivity. Tis in turn leads to increased production costs. Robots are ideal for monotonous work. Te speed of teir work is iger; tey are ceaper tan workers because people are inclined to fatigue. Tis is one of te reasons for te low prices of products. We propose anoter type of robots pedestrian, working at pedestrian crossings and performing a set of cyclic operations. In te real work one more type of collectives of robots transport, working in SRN of te city and cyclic operations carrying out a set is offered.. Robots in Transport System Te last twenty years te idea of management of motor transport witout participation of te driver actively develops. One of te directions of development of intellectual transport is development of veicles for intercity transportation of passengers. A distinctive feature of suc veicles is interaction of te passenger wit a veicle and absence of te operating person. Furter development in tis direction will allow creating intellectual information system of movement of veicles. Te last 0 years te idea of creation of transport witout participation of te driver (autonomous transport) actively develops. In view of wat te question of movement regulation of pilot fewer veicles is actually. One of te directions of intellectual transport development is integration autonomous veicles in intercity transportation of passengers. Distinctive feature of suc veicles is interaction of te passenger wit te veicle and lack of te operating person. Furter development in tis direction will allow creating intellectual information system of veicles movement. Te project Te robotics pipeline system Crosswalk [0] is known. Tis system carries out suttle transportations on a crosswalk of pedestrians troug te igway wit eavy motor transportation traffic. Te basis of system is made of te mobile robot. Electric veicle runs along a pedestrian crossing non-constantly. Transport detectors measure te intensity of traffic flow and in a case of falling below set value, te electric car moves to left side (Fig. ). In tis case, te crosswalk is used in te traditional way. Te electric car stays in parking place at nigt, and, as specified above, during te low intensity of traffic flows. At tis time e connects to te carging station. Figure. Te crosswalk system electric motor car at te moment of loading 40

45 Transport and Telecommunication Vol. 4, No, 03 It is te first type of te robots working in SRN of te city. In tis work one more type of te robots, wic are working also in SRN of te city and carrying out a set of cyclic operations of transportation of passengers, only any more across te igway, and along is offered and investigated. Te main objective of consideration is development of model of te automated system of transportation wit use of artificial intelligence. 3. Intellectual Transport Systems Intellectual transport systems (ITS) a class of te transport systems providing steady, effective, economic and safe functioning of transport at te expense of giving to active elements of transport system of properties of adaptive (intellectual) beaviour. Two most important aspects of intellectual transport system (ITS) are ATMS (Advanced Traffic Management Systems te advanced control system of movement) and ATIS (Advanced Traffic Information Systems te advanced information system) []. ITS a complex of te interconnected automated systems solving problems of management of traffic, monitoring and management of work of all means of transport (individual, public, cargo), and te enterprises on te organization of transport service in te region territory. Te automated transport systems, for wic te sort Englis-speaking term people-mover is tougt up, appeared in 967 in te form of te trains serving visitors of te World Fair in Montreal. Driverless car (te car witout te driver or an independent veicle) is a veicle equipped wit system of te autopilot, and capable to move from one point to anoter witout te aid of te operator. EC EUREKA Prometeus Project, gettere from te Neterlands, te researc company ARGO in Italy, management of te perspective researc DARPA programs in te USA, and Google driverless car are engaged in development of programs for autonomous cars. Te istory of independent veicles begins from 977 wit laboratory of mecanical engineering of Tsukuba in Japan. In te 980t years te video operated robot van of te Mercedes-Benz company developed by Ernst Dikmanns and is team at te Bundeswer University in Munic, reaced 00 km/ on streets witout movement. Subsequently, te European commission began financing wit 800 million euros te EUREKA Prometeus Project on independent veicles ( ) []. Also in te 980t years Te program independent land veicles (Autonomous Land Veicle ALV), financed by Management of te perspective researc DARPA programs of te United States, acieved te first demonstration on te road of an independent veicle wit use of a laser radar. In 994 two VaMP and Vita- cars robots from Daimler-Benz and Ernst Dikmanns from UniBwM overcame more tan one tousand kilometres on te Parisian tree-lane igway wit usual dense movement at a speed of 30 km/, toug tey were semi-autonomous wit uman intervention. Tey sowed independent driving in free lanes, driving in a column, te left and rigt transference on strips wit an independent pass aead of oter cars. In 995, converted independent Mercedes-Benz of an S-class Dikmanns made 600 kilometres travel from Munic to Copenagen and back, using faltering computer supervision and te so-called transputer tat is te computer working in real time. Te robot reaced speeds more tan 75 km/ on te German igway (wit te 95t % independent driving). In 995 te Navlab project (Carnegie Mellon's university) reaced independent driving for 98.%, aving overcome 5000 km. Tis car, owever, was semiautonomous by te nature: it used neural networks to operate a weel, but a trottle and a brake were operated by person. From 996 till 00 Alberto Brogki from university of Parma began te SLANG project wit te purpose of riding along marking of te igway. Te way in 000 km became te culmination of te project on igways of nortern Italy wit average speed of 90 km/. 94% of time te car was completely in an automatic mode. Te veicle ad only two black-and-wite ceap video cameras on a board and used tree-dimensional algoritms of supervision for environment recognition. In 008 of General Motors declared tat tey will start to test independent cars to 05, tat to 08 tey could be on te road. From te executed consideration it is possible to draw te following conclusions. Works in te field of creation of independent veicles will constantly accrue. In tis direction it is possible to allocate four approaces: - completely independent veicles; - various connections to infrastructure (eiter all area, or certain streets) to create te independent closed system; - systems of elp, wic by development on reduce requirements to te person driver (for example, updating of system of automatic control of speed or cruise control); - creation of road trains. 4

46 Transport and Telecommunication Vol. 4, No, 03 Wat is te development of system of road trains? To use road space better, veicles are collected in columns similar to a train were te driver (eiter uman or automatic) of te first veicle makes all decisions for all column. All oter veicles simply follow te leader te first veicle. For completely autonomous driving it is necessary, tat te car drives itself to te set purpose, using not canged infrastructure. Te ultimate goal in safe moving in any situation neverteless isn't reaced yet. Terefore it is possible to draw a conclusion tat at tis stage of development it is necessary to use as muc as possible possibilities of robots to perform intended work well specially in limited environment or in te environment wit not canged infrastructure. To suc systems belongs te automated transport system Crystal Mover from Mitsubisi wic as many interesting decisions and innovative development. Its cars in lengt of,84 m contain 03 passengers, and develop till 80 km/! In a track tey are kept by lateral rollers, and noise is cancelled wit car tires. Te driving gear is electric, a current moves troug te current collector, like in te subway. Tis system works not only at te airports Miami, Wasington, Hong Kong, Singapore, Seoul, but also in te cities Kobe, Tokyo, Hirosima, Kanazava, Singapore. Also to tis class of te systems working in specially prepared, limited environment or in te environment wit not canged infrastructure belongs automated transport system of mass conveyor transportation of passengers on te basis of mobile robots of development of te Brest state tecnical university (BrSTU). Intellectual transport system BrSTU is te complex system providing effective, economic, steady and safe functioning of transport at te expense of giving to elements of transport system of properties of adaptive, intellectual beaviour. Te intellectual information transport system of mass conveyor transportation of passengers is te difficult complex intended for optimisation of transportation of passengers in city boundaries in an independent mode, i.e. witout te driver. It is intended for transportation of passengers in Minsk. Te ig urgency of development consists in tat in Minsk currently on routes of CPT tere is lack of 500 drivers. 4. Description of Transport Intellectual System BrSTU Te model of representation of eac element is defined by its function in transport system. Te basis of system is represented by te following elements: intellectual stop; intellectual traffic ligt; te intellectual electric truck for transportation of passengers; electronic ticket; software of transport intellectual system; server center. 4.. Intellectual Stop Te intellectual stop is intended for improvement of quality of service of passengers of public transport at te expense of installation of various means: te information display provides to passengers at stops wit operational and actual information; information about arrival time of transport; advertising information; providing passengers wit tourist information on te city; call of te dispatcer and services of emergency reaction; video surveillance modules are intended for control of offences, definition of degree of load of a stop. Data on te current load at a stop, and also all calls of transport are transferred in te server centre for processing. Also data can be requested directly by te electric truck. 4.. Intellectual Traffic Ligt Te intellectual traffic ligt includes te following modules: module of regulation of traffic video surveillance module 4

47 Transport and Telecommunication Vol. 4, No, 03 Traffic is regulated so tat to provide te minimum idle times of te electric truck in case of te increased load. Data on te current load on a traffic ligt and about a condition of work are transferred in te server centre for processing. Also data can be requested by te electric truck Te Intellectual Electric Truck for Transportation of Passengers Te intellectual electric truck is intended for improvement of quality of service of transported passengers by means of te following modules: te module of sale of tickets for passengers at wic isn't present te electronic ticket manual purcase of te ticket in salon; te module of control of payment of journey for passengers at wic is te electronic ticket; te video surveillance module beind salon for control of offences (counts up number of passengers in salon, defines te violators wo ave not paid for journey); te module of satellite navigation for control of location; te wireless module of interaction wit oter electric trucks for realization of a multi-agent approac between electric trucks; te wireless module of interaction wit intellectual traffic ligts (te electric truck can request from a traffic ligt providing conditions of te minimum delays on traffic ligts in case of ig load); button of an emergency call of te dispatcer. Electric trucks excange information about te current load of stops, number of transported passengers. It provides realization of a principle of te multi agent system, allowing minimizing a waiting time of passengers at stops. Te passenger can cause te electric truck at a stop by means of te terminal. Te inquiry goes on te server centre wic on te basis of tese and oter data selects te optimum electric truck and sends it inquiry. Te electric truck wen receiving inquiry is obliged to stop at te specified stop and to take away passengers. In a case wen te amount of passengers is more tan empty seats in te electric truck, te inquiry in te centre on receiving one more electric truck is sent. After landing te passenger sould eiter get te ticket, or use previously te bougt electronic ticket Electronic Ticket Te ticket is executed in te form of analogue of a plastic credit bank card; provides replenisment of money on te ticket account, and as teir automatic removal at journey in te electric truck. Software of intellectual system includes: information processing from video cameras; processing of navigation information; interaction protocols between electric trucks and intellectual traffic ligts; collecting analytics and statistics and protocols of protection of information; electronic tickets; information displays and terminals Server Centre Te server centre represents complex control centre of transport intellectual system. It is presented by processing and data storage servers (economic and statistical information), realizing te following functional: information processing from video cameras at stops processing of requests from terminals at stops management of information displays information processing from video cameras in electric trucks information processing from intellectual traffic ligts analysis of load of routes and coice of optimum algoritm of work. 43

48 Transport and Telecommunication Vol. 4, No, Conclusions Te project offered in te article Mobile Autonomous robots a new type of city public transport discovers a new type of public transport and intends for increase of mobility and flexibility of transportation of passengers. Te project as a large economic profit. As a result of working at te project we ave considered two types of mobile robots tat are intended for functioning in road network of a city. Teir functioning takes place in confined environment wit distinct infrastructure. Possible fortuitous departures from te described environment are included in a zone of robot s reaction. On te base of tese robots te conception of new type of public transport is formed up. Tis project is in te stage of te draft. At present computer modelling of te system in different modes of intensity of passenger traffics is carried out. In te future autors ope for joint working wit suc an autoritative in Belarus producer of tram and trolleybus macinery as Belkommunmas. References. U. S. Department of Transportation Federal Higway Administration (FHWA). Retrieved June 0, 0, from ttp:// Kapitanov, V. T. (985). Management of transport streams in cities. V. T.Kapitanov, E. B. Kilazev (Eds.). M.: Transport. 3. Mikaylov, A. Yu. (004). Current trends of design and reconstruction of street road networks. A. Yu. Mikaylov, I. M. Golovnyk (Eds.). Novosibirsk: Science. 4. Anfilets, S., Sut, V. (00). Evaluating te Effectiveness of te Adaptive Control System in Brest Region. In Proceedings of te International Congress of Heavy Veicles, Road Trains and Urban Transport, 00, (pp. 6). Minsk: BNTU. 5. Kasianik, V. V., Anfilets, S. V., Suts, V. N. (00). Application of Artificial Neural Networks for Forecasting of Caracteristics of Transport Stream and Adaptive Regulation at Crossroads. In Proceedings of te International Conference on Neural Networks and Artificial Intelligence, 00 (pp. 9 93). Brest: BSTU. 6. Toyota s Humanoid Robot Violinist. Retrieved 009, from ttp:// 7. Robot susi. Retrieved 009, from ttp:// 8. Lovin, J. (007). We create te robot an android te ands. Moscow: DMK-Press. 9. Golovko, V. A. (00). Basis of artificial intelligence. V. A. Golovko, L. P. Matyuskov, V. N. Sut (Eds.). Brest: BrSTU. 0. Sut, V. N. (0). Te robotics Crosswalk pipeline system. Artificial Intelligence Journal, 3, Intellectual transport system. Retrieved 0, from ttp://scipeople.ru/group/435/.. Group of companies (group of companies) MM telematics. Retrieved 0, from ttp://mm-t.ru/. 44

49 Transport and Telecommunication Vol. 4, No, 03 Transport and Telecommunication, 03, Volume 4, No, Transport and Telecommunication Institute, Lomonosova, Riga, LV-09, Latvia DOI 0.478/ttj INDIVIDUAL ATTITUDES AND SHOPPING MODE CHARACTERISTICS AFFECTING THE USE OF E-SHOPPING AND RELATED TRAVEL Federica Crocco, Laura Eboli, 3 Gabriella Mazzulla University of Calabria Faculty of Engineering, Department of Land Use Planning P. Bucci, cubo 46/B, Rende (CS), Italy P Fax federica.crocco@unical.it P Fax laura.eboli@unical.it 3 P Fax g.mazzulla@unical.it New telecommunication tecnologies and services ave caused important callenges on travel beaviour and trip caracteristics. Existing literature studies sow many different outcomes regarding te effects of te new tecnologies on te participation of people to teir personal activities and related travel; specifically, e-sopping may produce te reduction of sopping trips (substitution effect), but also an increase of trips tanks to te reuse of te travel time saved for oter activities and trips (complementarity s effect). Te focus of tis study is to analyse te aspects mostly affecting consumer coices of purcasing goods by web or in-store, wit te aim of understanding ow to operate so tat e-sopping can positively modify consumers travel beaviour. Our researc findings sow tat individual social and economic factors, consumer attitudes, and sopping mode caracteristics influence te usage of online sopping. An experimental survey addressed to a sample of Italian consumers is used in te study. Keywords: e-commerce; travel beaviour; logistic regression models. Introduction and Motivation Demand for travelling is a derivative of te demand for performing activities. Everyday people move for conducting various activities, like working, studying, sopping, and so on. Terefore, people participation to te activities produces an amount of trips. Te penetration of Information and Communication Tecnologies (ICT) into uman life as influenced personal activities and related travel [,, 3, 4, 5, 6]. Te Internet, in fact, allows activities suc as working and sopping to be conducted witout travelling to te activity places. Te impacts of ICT may involve also some canges in land use, residential and work location, and possible transformations of social norms and values [7]. Many different activities can be made by ICT. As an example, Moktarian [8] classified eigt teleactivities: commuting, conferencing, sopping, banking, entertainment, education, medicine, and justice. However, te main personal activities tat can be conducted by ICT are telecommuting, telesopping, and teleleisure. Andreev et al. [9] sowed tat substitution was found as te most prevalent impact for telecommuting, wile complementarity is a major impact for telesopping and teleleisure. In te scientific literature, telecommuting as already received considerable attention, wile e-sopping and its relationsip wit personal travel beaviour as received far less attention. Most of te empirical studies consider e-sopping as part of ome-sopping (sopping by catalogue, television, fax, or telepone). However, e-sopping can be more properly defined as an activity to buy or to get information about consumer goods via te Internet; e-sopping also enables a spatial and temporal fragmentation and recombination of several stages in te sopping process. As an example, one can obtain information about a certain product by in-store sopping and purcase te product via te Internet; vice versa, it is also possible to use te Internet to obtain information about a product and ten buy it in a store [0]. As te popularity of sopping via te Internet increases, it could fundamentally cange people s travel beaviour, teir use of transport systems and te spatial configuration of te sops. Te expected benefit of e-sopping on transportation demand is te reduction of sopping trips (substitution effect) and consequently te possible reuse of te saved travel time for oter purposes/activities and trips (complementarity s effect) []. On te oter and, te combination of te increased access to product information wit variety via e-sopping, togeter wit te pleasure experienced by some individuals wile pysically sopping, may result in excess travel []. Te findings of Casas et al. [3] ad already sown tat online soppers do not travel less and in some cases travel more tan in-store soppers. Te study of Circella and Moktarian [4] confirms a int of complementarity of online sopping wit store sopping, consistently wit oter studies. Hjortol [5] 45

50 Transport and Telecommunication Vol. 4, No, 03 found tat tere is no indication of substitution of travel by sopping online and e-sopping seems to be a complementary activity to traditional sopping. Also, electronic commerce as a relevant influence on logistics; in fact, it is fundamentally canging te nature of supply cains, and redefining ow consumers learn about, select, purcase, and use products and services. In a conventional retailing, supply cain customers are responsible to purcase teir goods and services at te retailer s location. Given tat location is an important dimension of retailing, significant costs are assumed by te retailer to consider a location as accessible; te costs linked to te location are reflected in te final costs of a good. Te introduction of e-commerce as significantly canged te relationsip between customers and retailers (e-retailers). Te result as been te emergence of new business-to business supply cains tat are consumer-focused rater tan product-focused. Customers are directly linked to te supply cain since teir action of ordering a product reaces directly te distribution centre. In addition, customers want to know teir landed cost of te products, wen and were te products can be delivered, and tis callenge te distribution industry to implement information systems to track parcels as well as veicles. Definitively, e-commerce impacts supply cain management in a variety of keyways, suc as: cost efficiency (in terms of reduction of costs, improvement of data accuracy, enancement of customer service); distribution system (more flexibility in managing te movement of products and information between businesses, suppliers and customers); customer orientation (in terms of delivering better services to teir customers); sipment tracking (e.g. providing to users an account for obtaining real-time information about cargo sipments); sipping notice (e.g. receiving process by electronically transmitting a packing list aead of te sipment); sipping documentation (automatic production allows reduction of manual intervention); online sipping inquiry (in terms of instant sipping information access to anyone in te company, from any location). From a transportation perspective, we can igligt tat in te traditional system, te sopper was bearing te costs of moving te goods from te store to ome, but wit e-commerce, tis segment of te supply cain as to be integrated in te freigt distribution process. Te result potentially involves more packaging and more tons-km of freigt transported, especially in urban areas. Traditional distribution systems are tus ill fitted to answer te logistical needs of e-commerce. We can anticipate less use of private cars, but more employment of delivery vans, as well as smaller orders to be sipped longer distances. Te contradicting results of te various researc studies regarding te impact of e-sopping on travel imply te need of more researc on tis topic. Tere ave not been many empirical studies attempting to establis te relationsip between some factors and te intention to use or actually use online sopping. In te period after te downturn of te dot-com businesses, it became even more important to analyse tese factors [6] in order to understand consumers beaviour and perceptions about te opportunity of sopping online, wic could conduct to a reduction of te trips made to pysically reac te stores or at least to a more efficient organization of te trips. In tis study, te aspects mostly affecting user coices to use te Internet for purcasing goods or to buy in-store ave been analysed, wit te aim of understanding ow to operate so tat e-sopping can positively modify travel beaviour of te consumers. Te study is supported by a survey, wic provided experimental data collected from a sample of online and in-store consumers living in te Sout of Italy. Te next section provides a literature review igligting te influences of several factors on te sopping mode coice. In section 3 te experimental survey is depicted, wile section 4 reports te results of a statistical analysis of te data collected from te survey. Specifically, consumer and product caracteristics, and sopping mode features ave been analysed; te statistical analyses ave been carried out in order to define te profiles of te typical online and in-store consumer by analysing separately all te caracteristics of te two categories of consumers. A logistic regression model, reported in section 5, as been proposed in order to determine te variables explaining te propensity of people to online sopping. Finally, in a conclusive section te main considerations emerged from te researc ave been discussed.. Factors Affecting E-Sopping: a Literature Review Sopping is a process composed of a set of distinct components linked togeter in a particular sequence [7]. Moktarian [0] stated tat typical elements of te process include desire, information gatering/receiving, trial/experience, evaluation, selection, transaction, delivery/possession, display/use, and return, e.g. [8, 9]. Te coice of sopping mode can play a role in eac element of te sopping process [0]. Tere are various elements affecting coice between online and in-store sopping mode. Some literature studies analyse tese elements. Cang et al. [6] proposed a review arisen from te analysis of 45 articles, wit te aim of classifying te factors affecting te intention to use or actually use online sopping. Some factors were classified into tree main categories: perceived caracteristics of te web as a sale cannel, caracteristics of te consumers, and caracteristics of te website or products. Eac category includes subcategories. Te first category includes variables concerning te risk of online sopping perceived by consumers, relative advantages of online sopping, aspects regarding online 46

51 Transport and Telecommunication Vol. 4, No, 03 sopping experience, service quality factors, and trust of online transactions. Te category concerning te caracteristics of te customers includes consumer sopping orientations, demograpic variables, computer/internet knowledge and usage, consumer innovativeness, psycological variables. Te website and product caracteristics include risk reduction measures, website features like design, product caracteristics like cost and tangibility. Te factors analysed in te various studies sowed significant (positive or negative) or no significant impacts on te use of online sopping. Some studies verified tat te perceived risks concerning system security, credit card, and product ave a negative impact on te intention and usage of online sopping, wile variables representing relative advantages, like time saving, convenience, ease of use, utility as communication and distribution cannel, ave a positive impact on online sopping, as well as te variables of service quality, like customer service and perceived quality of e-vendors. Consumer caracteristics ave generally positive impacts on te online sopping. Examples of variables wit positive impacts are: convenience oriented, recreational oriented, impulsiveness, education level, age, income level, access to credit card. In addition, variables concerning computer/internet knowledge and usage ave a positive impact. Psycological factors suc as attitude, subjective norm, perceived beavioural control positively affect te intention and usage of online sopping. Also Farag et al. [] proposed a literature review of factors affecting e-sopping. Tey describe te classification of Salomon and Koppelman [9, ] wo identify four categories of caracteristics affecting te beavioural coices in te sopping process: sopping motive; product caracteristics; sopping mode caracteristics; individual caracteristics. Sopping activities ave several functions: economic functions, suc as buying a product; social functions (meeting people, conversation); recreational functions (pysical exercise, entertainment); psycological needs (exposure to information or to fres stimuli). Also product caracteristics can affect sopping mode coice. As an example, te products popularly purcased via e-sopping are computer ardware and software, CDs, books, travel tickets, cinema and concert tickets [3, 4, 5]; on te oter and, te products usually bougt in-store are clotes, furniture, and cosmetics. Concerning sopping mode caracteristics, e-sopping is rated relatively low in comparison wit in-store sopping on product information, product sales, security of transactions, and ease of returning mercandise [4, 6, 7]. In addition, e-sopping is rated relatively ig on timesaving and flexibility in sopping ours. Individual caracteristics also affect sopping mode; as an example, e-sopping is mainly done by young male graduates in professional occupations and wit ig incomes [3, 8, 9]. Dijst et al. [30] found tat people wit more Internet experience ave a stronger inclination to buy online; moreover, aving a credit card positively affects te volition to buy a media product online. On te oter and, wit respect to te in-store sopping, people wo own one or two cars ave a relatively weak volition to buy a media product. Earlier researc as found tat owners of one or two cars make fewer sopping trips tan people wo do not own a car [3, 3]. Hsiao [0] proposed a literature review on te attributes of sopping mode classified into tree main categories: information gatering/sopping, purcase/transaction, delivery. According to Salomon and Koppelman [9], some of te attributes may serve economic function, and some psycological function. Te category of information gatering/sopping includes caracteristics concerning travel; specifically, in-store sopping involves travel cost and time. Oter caracteristics regard psycological factors linked to sopping fun and information uncertainty; specifically, online sopping involves less sopping fun and more information uncertainty tan in-store sopping. Te category of purcase/transaction includes caracteristics concerning purcase price and distrust feelings caused by transaction. Te category of delivery includes waiting time for delivery and oter inconveniences caused by delivery. 3. Survey Te study area is te single built-up area made up of te towns of Cosenza and Rende, sited in te Sout of Italy. Te urban area as grown over te years also tanks to te presence of te University of Calabria, wic expanded in te Nort of Rende at te beginning of te seventies. Cosenza and Rende represent a centre of attraction for all te towns of te wole province, because of te administrative functions, job opportunities and supply of services. Te urban area as about 0,000 inabitants. In addition, a fair proportion of university students from oter places of Calabria lives in Rende or Cosenza; te students attending te University of Calabria are approximately 35,000. Te total number of employees in te urban area is about 37,000. About 47% of te employees work in te public sector, about 37% in te business or te oter private services, only 4% in te industry and % in te agriculture. A web survey was conducted in order to investigate te factors affecting te inclination of people to te usage of e-sopping compared to in-store sopping. Te survey was realized at te end of 008 by te website of an online magazine publised by te University of Calabria. About 3,400 frequent readers of te magazine were reaced by , and,6 individuals voluntarily took part in te survey; 47

52 Transport and Telecommunication Vol. 4, No, 03 terefore, te response rate is 35.8%. Eac respondent was asked for describing is/er last purcase by indicating te category of product purcased and te adopted sopping mode (in-store or online). On te basis of tis last statement te sample was divided into two groups, by assuming tat wo adopted online sopping mode is an online consumer, wile wo purcased in a store is an in-store consumer. Terefore, te sample is divided in by 56 online consumers and 654 in-store soppers. Te questionnaire was divided into tree sections. Section aimed to collect information about some socio-economic caracteristics of te sample (gender, age, employment, education, income, and so on). Section aimed to collect information about te purcase (category of te product, price of te product, sopping mode). Some information about consumer attitudes for sopping are collected troug section 3. Specifically, consumers were asked for giving a judgement according to a five-point verbal scale ( irrelevant, little important, quite important, very important, fundamental ) to teir following attitudes: need of andling and aving a close look at products; need of interacting wit sop assistants; propensity to negotiate; inclination to make recreational activities for sopping; propensity to purcase witout travelling to sopping places; perceived risk concerning credit card; perceived problems concerning prepayment. Section 3 contains also some questions in wic consumers were asked for giving a judgement according to te same verbal scale to factors concerning tree peculiarities of in-store sopping and fifteen caracteristics of online sopping: presence of a pysical structure; opportunity of immediately aving products; easiness to cange products; slowness of te Internet connection; large variety of offered products; opportunity of acquiring more in detail information about products; opportunity of comparing prices and saving; opportunity of gaining information from oter people netsurfing; opportunity of purcasing at any time; opportunity of safely purcasing by avoiding robberies; opportunity of purcasing articles unobtainable on traditional cannels; opportunity of aving online assistance in real time; opportunity of using digital cas; opportunity of using wallet; problems concerning dispatcing; problems concerning delivery; opportunity of taking advantage of discount. Out of,6 respondents, tere are 666 males (55%) (Table ). Most of te sample is between 4 and 65 years old (45.8%); relevant percentages are reported for people between 8 and 4 (9.9%) and between 5 and 40 (9.5%). Half of te respondents are employees and 34% are students. Over alf of te employed people are clerks and 3% are freelancer. Most of te employees work in te sector of public administration (56.7%). Over alf of te sample as a degree. Half of te sample belongs to an uppermiddle class of income, and 33.3% of te respondents belong to a lower-middle class. Te classes of income refer to te net montly income of te family unit, expressed in Euros. Te number of members in a family unit is 3.6 on average. Eac family as.6 members wit car driving licence and.9 cars on average. Te z-test for a proportion was conducted to compare te proportions created by te random sample to te proportions in te population in order to determine te representativeness of te sample. Since te test statistic as an absolute value iger tan te critical value of.96 (level of significance of 5%), or.645 (level of significance of 0%), te researcer rejects te null ypotesis tat tere is a statistically significant difference between te population proportion and te sample proportion. Te obtained test values suggest tat te sample can be considered as representative of te population. 4. Sample Statistical Analysis 4.. Consumer Caracteristics A more useful descriptive statistical analysis was carried out in order to define te profiles of te online and in-store consumer by analysing separately te socio-economic caracteristics of te two categories of consumers (Table ). As above mentioned, te sample is made up of 56 online consumers (46.%) and 654 in-store soppers (53.8%). Little more tan alf of in-store consumers are females, wile 55.% of online consumers are males. From te responses given by te sample, te following profile of te online consumer emerges: online consumer is a university or secondary scool student (55.0%), wit an age between 8 and 4 years (47.7%), and a iger-middle level of family income (48.6%). Terefore, tis statistical analysis suggests tat being male and young, being student and aving a ig family income positively increase te propensity to buy online. A profile of in-store consumer can be described according to te same criteria: in-store consumer is an employee (65.%), belonging to a iger-middle class of income (44.0%) and aving an age between 4 and 65 years (76.3%). As expected, in-store consumer as more possibility to use te private car, given tat 8.7% of tese consumers belong to families aving 3 or more cars (against 4.3% of te online consumers), and 65.4% of tem belong to families wit 3 o more members aving car driving licence (against 50.4% of te online consumers). Te two-proportion z-test was conducted to determine weter te difference between te percentages of te two samples (or te two categories of consumers) is significant. Te z-test verifies 48

53 Transport and Telecommunication Vol. 4, No, 03 from te observed sample difference weter te ypotesized difference between population proportions differs significantly. If te sample findings are unlikely, given te null ypotesis tat te two proportions are equal, te researcer rejects te null ypotesis. Specifically, te null ypotesis is rejected if te absolute value of te statistics is iger tan.96 (level of significance of 5%), or tan.645 (level of significance of 0%). By observing te obtained values of te z-test reported in table, we can say tat te two populations differ significantly, except for belonging to te category of workers, compulsory education, upper and lower-middle income classes. Te test is not satisfactory also for some caracteristics regarding te family size, driving licence and car ownersip. Table. Frequency of te responses about consumer socio-economic caracteristics (n =,6) Caracteristics In-store consumers (654) Online consumers (56) Two proport ion z- test Total (6) Gender male 356 (54.4%) 30 (55.%) (55.0%) female 98 (45.6%) 5 (44.8%) (45.0%) Age < 8 8 (.%) 45 (8.0%) (4.4%) (4.7%) 68 (47.7%) (9.9%) (7.0%) 9 (34.0%) (9.5%) (76.3%) 58 (0.3%) (45.8%) > 65 5 (0.8%) 0 (0.0%).08 5 (0.4%) Employment employed 46 (65.%) 90 (33.8%) (50.7%) unemployed 3 (4.7%) 44 (7.8%) (6.%) searcing for a first emp. 7 (.%) 3 (.3%) -.70* 0 (.6%) ousewife 56 (8.6%) 6 (.%) (5.%) ig scool student 8 (.%) 45 (8.0%) (4.4%) university student 98 (5.0%) 64 (47.0%) (9.8%) pensioner 8 (4.3%) 0 (0.0%) (.3%) Career clerk 6 (53.%) 5 (65.8%) -.95* 35 (57.0%) manager 0 (4.7%) 8 (4.%) n.s. 8 (4.5%) entrepreneur 65 (5.3%) 0 (5.3%) 3.50* 75 (.%) freelancer 05 (4.6%) 37 (9.5%) n.s. 4 (3.%) artisan 6 (.4%) 0 (0.0%).65* 6 (.0%) worker 3 (0.7%) 4 (.%) n.s. 7 (.%) oter (0.%) 6 (3.%) -3.6* 7 (.%) Education compulsory education 5 (7.8%) 50 (8.9%) n.s. 0 (8.3%) ig scool leaving qualifications 48 (.6%) 67 (47.5%) (34.%) degree 4 (64.5%) 3 (4.%) (53.7%) post degree qualification 33 (5.0%) 4 (.5%) (3.9%) Family income level < 0,000 euros 0 (8.3%) 9 (4.4%) (.3%) 0,000-40,000 euros 6 (33.0%) 89 (8.9%) n.s. 405 (33.3%) 40,000-60,000 euros 88 (44.0%) 38 (48.6%) (49.8%) > euros 30 (4.6%) 6 (4.6%) n.s. 56 (4.6%) Family members member (0.3%) 0 (0.0%) n.s. (0.%) members 33 (5.0%) 35 (6.3%) n.s. 68 (.8%) 3 members 36 (36.0%) 5 (40.0%) n.s. 46 (37.9%) 4 or more members 383 (58.6%) 30 (53.7%).69* 685 (56.3%) Family members wit member 6 (.5%) 6 (.%) n.s. 0 (0.0%) car driving licence members 0 (3.%) 73 (48.5%).80* (.8%) 3 or more members 48 (65.4%) 83 (50.4%) (58.5%) Family car ownersip 0 3 (0.5%) 0 (0.0%) n.s. 3 (0.%) 74 (.3%) 99 (7.6%) (4.%) 50 (79.5%) 439 (78.%) n.s. 959 (78.9%) 3 or more cars 57 (8.7%) 4 (4.3%) (6.7%) *significant at a level of 0% sample Also concerning consumer attitudes, a first analysis can be conducted by observing for eac question te frequency of response (Table ). Considerations tat are more useful can be made if te data are analysed by distinguising te responses of people wo cose in-store sopping from te responses of people wo purcased online. Tis type of analysis elps to better and more immediately understand factors influencing sopping mode coice. For most of online consumers (68%) te need of andling and aving a close look at goods is irrelevant or little important; vice versa, 64% of in-store consumers considers as fundamental tis kind of need. For almost all in-store consumers (88%) te inclination to make recreational activities for sopping is fundamental; wile online consumers ave conflicting opinions, given tat a relevant percentage of tem (4%) retains tis aspect as little important and a percentage as relevant (44%) considers te aspect 49

54 Transport and Telecommunication Vol. 4, No, 03 as fundamental. Concerning te propensity to interact wit sop assistants, most of people wo purcased online (6%) give little or no importance to tis aspect, wile 65% of in-store consumers consider as important tis aspect. An unexpected outcome is obtained for te factor regarding te propensity to negotiate; specifically, most of people wo purcased in-store (68%) retains tis aspect as irrelevant or little important (even if in tis last case te z-test is not satisfactory), wile over alf of online consumers consider tis aspect as important. In tis case, te propensity to negotiate reflects a need of te online consumers to competitively priced purcase and to coose te sopping mode, wic generally guarantees lower prices. Te opportunity to purcase witout travelling to sopping places is considered as very important or fundamental by almost 60% of bot te categories of interviewed consumers. An expected outcome was obtained for te aspect linked to te perceived risk concerning credit card; specifically, 9% of online consumers considers tis risk as little or no important, wile only 3.5% of in-store consumers does not give importance to tis risk. Te aspect linked to te perceived problems concerning prepayment does not sow particular differences by comparing responses expressed by people wo purcased online wit te responses of consumers wo purcased in-store. Table. Frequency of te responses about consumer attitudes irrelevant little important quite important very important fundamental andling and aving a close look in-store consumer (%) online consumer (%) two proportion z-test interacting wit sop assistants in-store consumer (%) online consumer (%) two proportion z-test * negotiate in-store consumer (%) online consumer (%) two proportion z-test 7.79 n.s. n.s recreational activity in-store consumer (%) online consumer (%) two proportion z-test purcasing witout travelling in-store consumer (%) online consumer (%) two proportion z-test 4.9 n.s. n.s risk concerning credit card in-store consumer (%) online consumer (%) two proportion z-test problems concerning prepayment in-store consumer (%) online consumer (%) two proportion z-test -5.9 n.s. n.s *significant at a level of 0% In addition, a question about consumer experience wit te new tecnologies was addressed to all te users. Specifically, 97% of online consumers state to ave experience wit online sopping. In-store consumers, instead, are less expert wit new tecnologies; in fact, over alf of tem (54%) states to ave not experience wit online sopping. Finally, additional questions were addressed to te in-store consumers about teir potential usage of te Internet before purcasing. Specifically, 73.4% of te consumers used te Internet for collecting information about te products before purcasing in-store; almost 60% of te consumers used te Internet for searcing for te different articles or brands of te desired product and controlling te product prices; about 30% of users used te Internet for searcing for te most convenient point of sale nearest to ome or more convenient in terms of assortment and promotion, and only 9% for searcing for information from people experienced at buying te desired products. 4.. Product Caracteristics Eac respondent specified te category and te level of price of te product tat e/se purcased (Table 3). Te products were preliminary classified according to sixteen categories, wile nine categories of price were predefined. Altoug about 30% of bot online and in-store consumers purcase products of electronics and telepony, online consumers purcase more products of computer ardware and notebook (3.5%), wile in-store consumers purcase a certain quantity of articles of cloting (8.9%). Tere are many categories 50

55 Transport and Telecommunication Vol. 4, No, 03 of products purcased by scant percentages of consumers; for tese categories, te values of te two-proportion z-test are not satisfactory. Over alf of te products purcased online ave a price between 300 and 500 Euros (55.9%), wile te majority of te products purcased in-store as a price between 50 and 00 Euros (40.7%). We can say tat online consumers tend to purcase products wit iger levels of price and wic can be easily cosen on te web, being products wic less need to be andled or ad a close look. Concerning te levels of price, we can say tat people purcase more online expensive products because tey can generally purcase on te ceap by web. On te oter and, in-store consumers tend to purcase larger quantities of product of cloting tan online consumers do, because tese kinds of products need more to be andled or ad a close look. Table 3. Frequency of te responses about product caracteristics Product category Product (Euro) price publising/books dvd/music computer ardware/notebook electronics and telepony computer software domestic appliances cloting soes personal accessories ealt sport perfumery toys jewellery ironmongery electrical materials lower tan or equal to 0 between and 50 between 5 and 00 between 0 and 300 between 30 and 500 between 50 and 700 between 70 and 900 between 90 and 000 iger tan 000 In-store consumers (654) (3.%) 47 (7.%) 74 (.3%) (33.8%) 0 (.5%) 40 (6.%) 89 (8.9%) 9 (.9%) 6 (.4%) 5 (0.8%) 9 (.4%) (0.3%) 0 (0.0%) (0.%) 0 (0.0%) 0 (0.0%) (0.%) 57 (4.0%) 66 (40.7%) 37 (0.9%) 4 (.%) 3 (4.7%) 9 (.9%) 9 (4.4%) 0 (0.0%) Online consumers (56) 0 (0.0%) 40 (7.%) 77 (3.5%) 60 (8.5%) 0 (0.0%) 78 (3.9%) 3 (5.5%) 9 (3.4%) 3 (4.%) 9 (3.4%) 6 (.%) 7 (.%) 0 (0.0%) (0.4%) 0 (0.0%) 0 (0.0%) 35 (6.%) 99 (7.6%) (.0%) 48 (8.5%) 34 (55.9%) 4 (4.3%) 7 (.%) 4 (4.3%) 0 (0.0%) Two proportion z-test 4.9 n.s n.s. n.s n.s. -.9 * n.s. n.s. n.s. n.s n.s..99 n.s. n.s. Total sample (6) (.7%) 87 (7.%) 5 (0.6%) 38 (3.3%) 0 (0.8%) 8 (9.7%) 0 (8.%) 38 (3.%) 39 (3.%) 4 (.0%) 5 (.%) 9 (0.7%) 0 (0.0%) 3 (0.%) 0 (0.0%) 0 (0.0%) 36 (3.0%) 56 (.%) 77 (.8%) 85 (5.%) 38 (7.0%) 55 (4.5%) 6 (.%) 53 (4.4%) 0 (0.0%) *significant at a level of 0% 4.3. Sopping mode Caracteristics All te consumers interviewed gave information about factors caracterizing in-store and online sopping, independently of teir coice about sopping mode. Terefore,,6 observations were collected for eac factor. Also for tese factors, te data are analysed by distinguising te responses of people wo cose in-store sopping from te responses of people wo purcased online (Table 4). Concerning te caracteristics of in-store sopping, tere are relevant differences in responses between people wo cose in-store sopping and people wo cose online sopping. Specifically, most of people wo purcased online (6%) give little or no importance to te opportunity of immediately aving products, wile most of people wo purcased in-store (86%) consider as fundamental, very or quite important tis kind of opportunity. As expected, te easiness to cange products is considered more important for people wo purcased in-store; in fact, a relevant percentage of online consumers (5%) consider tis aspect as little or no important. Finally, te presence of te pysical structure is judged rougly in te same way by bot te categories of consumers. By analysing te caracteristics of online sopping, not relevant differences in responses between people wo cose in-store sopping and people wo purcased online can be observed for many aspects. Specifically, te factors regarding te large variety of offered products and te opportunity of acquiring more in detail information about products are considered as fundamental by almost all people wo purcased online (84% and 76%, respectively), wile relevant percentages of people wo purcased in-store consider tese two aspects as little or no important (4% and 7%, respectively). Concerning te slowness of te Internet connection, te sample of people wo purcased online is divided between consumers considering tis aspect as irrelevant (57%) and consumers retaining it as a very important aspect (36%); on te oter and, 85% of people wo purcased in-store give little or no importance to tis 5

56 Transport and Telecommunication Vol. 4, No, 03 aspect. A series of aspects sows very similar percentages of response from te comparison between te rates expressed by te consumers belonging to te two different categories: opportunity of comparing prices and saving; opportunity of gaining information from oter people netsurfing; opportunity of purcasing at any time; opportunity of purcasing articles unobtainable on traditional cannels; opportunity of using digital cas; opportunity of using wallet; problems concerning delivery. Specifically, most of te online consumers (about 80%) give importance to te opportunity of comparing prices and saving; over 60% of te online sample considers te opportunity of purcasing at any time as a very important or fundamental aspect. However, te values of te z-test regarding tese two aspects do not allow to state tat tese differences are significant in te populations. Table 4. Statistical analysis of sopping mode caracteristics irrelevant little important quite important very important fundamental In-store caracteristics Online caracteristics *significant at a level of 0% presence of a pysical structure in-store consumer (%) online consumer (%) two proportion z-test * -.56 n.s. immediately aving products in-store consumer (%) online consumer (%) two proportion z-test n.s easiness to replace products in-store consumer (%) online consumer (%) two proportion z-test * slowness of te Internet connection in-store consumer (%) online consumer (%) two proportion z-test n.s n.s. large variety of offered products in-store consumer (%) online consumer (%) two proportion z-test acquiring more information in-store consumer (%) online consumer (%) two proportion z-test n.s comparing prices and saving in-store consumer (%) online consumer (%) two proportion z-test.50 n.s. n.s. -.6 n.s. gaining information from oter people in-store consumer (%) online consumer (%) two proportion z-test 4.69 n.s. n.s purcasing at any time in-store consumer (%) online consumer (%) two proportion z-test.37 n.s. n.s. n.s. n.s. safely purcasing in-store consumer (%) online consumer (%) two proportion z-test n.s n.s n.s. purcasing articles unobtainable on in-store consumer (%) traditional cannel online consumer (%) two proportion z-test 4.4 n.s * aving online assistance in real time in-store consumer (%) online consumer (%) two proportion z-test.9 * n.s taking advantage of discount in-store consumer (%) online consumer (%) two proportion z-test using digital cas in-store consumer (%) online consumer (%) two proportion z-test n.s..64 * -.89 * using wallet in-store consumer (%) online consumer (%) two proportion z-test n.s. n.s problems concerning dispatcing in-store consumer (%) online consumer (%) two proportion z-test * n.s. problems concerning delivery in-store consumer (%) online consumer (%) two proportion z-test n.s. n.s n.s..09 5

57 Transport and Telecommunication Vol. 4, No, 03 Te following factors do not sow particular differences by comparing responses expressed by people wo purcased online wit te responses of consumers wo purcased in-store: opportunity of safely purcasing by avoiding robberies; opportunity of aving online assistance in real time; opportunity of taking advantage of discount. As expected, te aspect linked to te problems concerning dispatcing is considered as more important for people wo purcased online. From te analysis of te frequency of responses, it emerges tat te in-store consumer as more inclination to make recreational activities for sopping, more need of andling and aving a close look at products, and of interacting wit sop assistants; at te same time, e/se retains important te opportunity of immediately aving products and te easiness to cange products. Instead, te online consumer retains tat te large variety of offered products, te opportunity of acquiring more in detail information about products, togeter wit te opportunity of taking advantage of discount are fundamental aspects of te online sopping; e/se also as more propensity to purcase witout travelling to sopping places, and considers as important te aspects linked to te opportunity of comparing prices and saving, and of purcasing articles unobtainable on traditional cannels. 5. Logistic Regression Models Te logistic regression tecnique as been adopted in order to determine te variables explaining te propensity of people to online sopping. Some consumer attitudes and sopping mode caracteristics ave been used as independent variables affecting sopping mode coice (Table 5). We introduced a binary logistic regression model for explaining online buying by te Internet users to control for te multivariate effects of te independent variables. Binary logistic regression predicts te probability to sop online compared to sop non-online. Te dependent variable of te model is te coice of online or in-store sopping. Te variable as a value of if te consumer purcased by web, and a value of 0 if te consumer purcased in a store; te value of 0 is te reference level representing te non-online sopping mode. Eac independent variable as a value of if te respondent judged te attitude or sopping mode caracteristic as fundamental, or very important, or quite important ; on te oter and, te 0 value corresponds to te little important or irrelevant judgement. A backward Wald procedure was used to determine significant predictor variables by using SPSS software. Backward selection starts wit all variables and removes one at a time, in te order tey are worst according to te Wald test. Te values of te statistics on te goodness of fit are reasonable. Specifically, R-square of Nagelkerke is 0.887, ci-square test is significant wit a value of , and te Hosmer-Lemesow test sows a ci-square value of wit a significance level of By observing te classification tables, te percentage of correct responses is 95.9% against 46.%, wic is te percentage wen tere are no variables in te model. Te results partly confirm te findings of te descriptive analysis. Te significant un-standardized logit coefficient for eac predictor in te model is sown in table 5, togeter wit its standard error, te Wald statistic, te p significance level, and te odds ratio. Table 5. Logistic regression analysis of online purcasing accounting for consumer attitudes and sopping mode caracteristics (n =,6) Coeff. β E.S. Wald Sig. Exp(β) interacting wit sop assistants ( = imp; 0 = not imp) negotiate ( = imp; 0 = not imp) risk concerning credit card ( = imp; 0 = not imp) immediately aving products ( = imp; 0 = not imp) large variety of offered products ( = imp; 0 = not imp) acquiring more information ( = imp; 0 = not imp) purcasing articles unobtainable on traditional cannels ( = imp; 0 = not imp) aving online assistance in real time ( = imp; 0 = not imp) taking advantage of discount ( = imp; 0 = not imp) Among te consumer attitudes, te need of interacting wit sop assistants and te perceived risk concerning credit card negatively affect te propensity to purcase online; wile te propensity to negotiate as a positive impact to te coice of online sopping. Among te caracteristics of te sopping modes, only te opportunity of immediately aving products as a negative impact on online sopping. All te oter sopping mode caracteristics introduced in te model positively affect te propensity to coose online sopping; specifically, te opportunity of taking advantage of discount as 53

58 Transport and Telecommunication Vol. 4, No, 03 te igest positive weigt on te coice of sopping online. From tese results we can conclude tat online consumer as more propensity to negotiate, e/se retains important aving discount by buying online, acquiring more information tan in a store, purcasing articles unobtainable on traditional cannels, aving a large variety of products in te Internet, and aving online assistance in real time. On te oter and, te in-store consumer perceives more te risk in te use of credit card concerning privacy wen e/se buys in te Internet, e/se as more need to interact wit sop assistants; in addition, instore consumer considers as important te opportunity to immediately ave products in-store. Odds ratios may be more usefully adopted to compare te relative strengt of te independents. By analysing te odds concerning te variables of te consumer attitudes, we can say tat te odds of online sopping compared to in-store sopping are decreased by a factor of 0.8 by needing interactions wit sop assistants; in addition, we could say tat te odds a consumer needing to negotiate purcases online are 3.04 te odds a consumer wo does not need to negotiate purcases online; finally, te probability of purcasing online, given tat consumers ave a ig perceived risk concerning credit card, is 0.00 te probability of purcasing online given tat tis risk is not considered by te consumers. Te effects of te variables concerning sopping mode caracteristics can be analogously explained. For te caracteristics concerning immediately aving products, te odds a consumer giving importance to tis aspect purcases online are te odds a consumer wo does not give relevance to tis aspect purcases online. In addition, we can say tat te odds of online sopping compared to in-store sopping are increased by a factor of 5.69 by considering as important te advantage of discount by web, 3.6 by giving importance to te possibility of acquiring more information by te Internet,.834 by retaining important te opportunity to purcase articles unobtainable on traditional cannels,.99 by giving importance to te large variety of products offered by web, and.73 by retaining important te online assistance in real time. 6. Conclusions Altoug many studies ave analysed te impacts of ICT on travel beaviour, different and contrasting outcomes were found. Tere is still vagueness regarding te effects linked to te interactions between ICT and travel beaviour, in terms of reduction of sopping trips (substitution effect) and te possible reuse of te saved travel time for oter purposes/activities and trips (complementarity effect). Researcers ave probably overestimated te importance of tecnologies and underestimated te importance of social factors, and te expected travel decrease as not been realized yet. Te proposed researc as tried to analyse te aspects mostly affecting user coices to purcase goods by te Internet or to buy in-store just for understanding ow e-sopping can positively modify te travel beaviour of te consumers. We ave analysed bot individual (socio-economic and attitudinal) and sopping mode caracteristics, troug statistical analyses of experimental data. Individual socio-economic caracteristics suc as being male and young, aving a ig family income affect positively te usage of online sopping. In addition, some online sopping mode caracteristics influence positively te use of te Internet for purcasing goods, suc as te large variety of offered products, opportunity of acquiring more in detail information about products, taking advantage of discount, comparing prices and saving, and purcasing articles unobtainable on traditional cannels. Online consumers tend to purcase products wit iger levels of price and wic can be easily cosen on te web being products wic less need to be andled or ad a close look. Logistic regression analysis largely confirms te above-mentioned outcomes. In addition, it emerges tat te attitude of te consumers to negotiate and te opportunity of aving online assistance in real time affect positively te usage of e- sopping. On te oter and, te typical in-store consumer is an employee, between 4 and 65 years old, working in te sector of public administration and belonging to a iger-middle class of family income. In-store consumer is recreational oriented, e/se needs to andle and ave a close look at goods, to interact wit sop assistants, and e/se perceives more te risk of te use of credit card concerning privacy wen e/se buys in te Internet. At te same time, e/se retains important te opportunity of immediately aving products and te easiness to replace goods. In-store consumers tend to purcase larger quantities of product of cloting. Logistic regression analysis confirms tese outcomes. Some outcomes of te proposed study are in line wit te findings of oter literature studies. As igligted from te literature review proposed by Cang et al. [6], we verified tat variables regarding perceived risk concerning credit card ave a negative impact on te usage of online sopping, wile variables representing relative advantages, like convenience, utility as communication and distribution 54

59 Transport and Telecommunication Vol. 4, No, 03 cannel, ave a positive impact on e-sopping. According to te same literature review, consumer caracteristics in general ave positive impacts on te online sopping, suc as education level, age, income level. In line wit te studies of Vrecopoulos et al. [3], Hjortol [8], Zmud and Arce [9], we found tat being young and aving a ig family income affects positively te usage of online sopping. Like Vrecopoulos et al. [3], Lee [4], and Szymanski and Hise [5], we found tat online consumers tend to purcase computer ardware and software, wile clotes are more purcased in-store. In line wit te study of Dijst et al. [30], we found tat people wit more Internet experience ave a stronger inclination to buy online. From te proposed studies, many oter factors concerning sopping mode caracteristics ave sown different effects on te usage of online sopping. Te outcomes of te proposed researc could be used for identifying te most convenient strategies to positively modify travel beaviour of te consumers. Some suggestions arise from te analysis of te caracteristics of bot online and in-store sopping mode caracteristics. As an example, a useful strategy could be adopted to encourage in-store soppers to acquire product information suc as prices or brands before purcasing in-store and to identify te most convenient points of sale. For tis aim, sopkeepers could provide online information regarding saleable products and store location to te consumers. Te acievements of sopping centres could be furter stimulated in order to satisfy user needs in doing bot sopping and oter recreational activities (cinema, sport, etc.). Tese strategies could elp to optimise te trips made by consumers to reac te sops. In addition, strategies for improving online sopping may be usefully adopted. As an example, one of te most relevant problems of te online sopping is caused by te door-to-door delivery of te products increasing te traffic of te delivery vans. In order to solving tis problem, specific delivery centres for te online consumers could be realized. References. Salomon, I. (986). Telecommunications and travel relations: a review. Transportation Researc Part A, 0(3), Salomon, I. (998). Tecnological cange and social forecasting: te case of telecommuting as a travel substitute. Transportation Researc Part C, 6, Handy, S. L. & Yantis, T. (997). Te impacts of telecommunications tecnologies on nonwork travel beaviour. Texas: University of Texas at Austin; Center for Transportation Researc. 4. Golob, T. F. & Regan, A. C. (00). Impacts of information tecnology on personal travel and commercial veicle operations: researc callenges and opportunities. Transportation Researc Part C, 9, Moktarian, P. L., Salomon, I. & Handy, S. (004). A taxonomy of leisure activities: te role of ICT. California: University of California, Davis; Institute of Transportation Studies. 6. Weltevreden, J. W. J. & Rotem-Mindali, O. (009). Mobility effects of bc and cc e-commerce in te Neterlands: a quantitative assessment. Journal of Transport Geograpy, 7, Salomon, I. (000). Can telecommunications elp solve transportation problems. In D. A. Henser & J. B. Kennet (Eds.), Handbook of Transport Modelling (pp ). Oxford: Pergamon. 8. Moktarian, P. L. (990). A typology of relationsips between telecommunications and transportation. Transportation Researc Part A, 4, Andreev, P., Salomon, I. & Pliskin, N. (00). Review: State of teleactivities. Transportation Researc Part C, 8, Moktarian, P. L. (004). A conceptual analysis of te transportation impacts of BC e-commerce. Transportation, 3, Ferrell, C. E. (005). Home-based Telesopping and Sopping Travel: Were do we find te Time? Transportation Researc Record, 96, 3.. Papola, A. & Polydoropoulou, A. (006). Sopping-Related Travel in a Ric ICT Era. A Case study on te Impact of e-sopping on Travel Demand. In Proceedings of te 85t Annual Meeting of te Transportation Researc Board, January 6, 006. Wasington DC. 3. Casas, J., Zmud, J. & Bricka, S. (00). Impact of Sopping Via Internet on Travel for Sopping Purposes. In Proceedings of te 80t Annual Meeting of te Transportation Researc Board, January 7, 00. Wasington DC. 4. Circella, G. & Moktarian, P. (00). Complementarity or Substitution of Online and In-Store Sopping: an Empirical Analysis from Nortern California. In Proceedings of te 89t Annual Meeting of te Transportation Researc Board, January 0 4, 00, Wasington, D. C. 55

60 Transport and Telecommunication Vol. 4, No, Hjortol, R. (009). Information searcing and buying on te Internet: travel-related activities? Environment and Planning Part B, 36, Cang, M. K., Ceung, W. & Lai, V. S. (005). Literature Derived Reference Models for te Adoption of Online Sopping. Information & Management, 4, Peterson, R. A., Balasubramanian, S. & Bronnenberg, B. J. (997). Exploring te implications of te internet for consumer marketing. Journal of te Academy of Marketing Science, 5 (4), Sciffman, L. G. & Kanuk, L. L. (987). Consumer Beavior, 3rd edition. Englewood Cliffs, NJ: Prentice-Hall, Inc. 9. Salomon, I. & Koppelman, F. (988). A framework for studying telesopping versus store sopping. Transportation Researc Part A, (4), Hsiao, M. H. (009). Sopping mode coice: Pysical store sopping versus e-sopping. Transportation Researc Part E, 45, Farag, S., Dijst, M. & Lanzendorf, M. (003). Exploring te use of e-sopping and its impact on personal travel beaviour in te Neterlands In Proceedings of te 8nd Annual Meeting of te Transportation Researc Board, January 6, 003, Wasington, D.C.. Salomon, I. & Koppelman, F. (99). Telesopping or Going Sopping: An Information Acquisition Perspective. Beaviour and Information Tecnology,, Vrecopoulos, A. P., Siomkos, G. J. & Doukidis, G. I. (00). Internet Sopping Adoption by Greek Consumers. European Journal of Innovation Management, 3, Lee, P. (00). Beavioral Model of Online Purcasers in E-commerce Environment. Electronic Commerce Researc,, Szymanski, D. M. & Hise, R. T. (000). E-Satisfaction: An Initial Examination. Journal of Retailing, 76, Koppelman, F., Salomon, I. & Proussalogou, K. (99). Telesopping or Store Sopping? A Coice Model for Forecasting te Use of New Telecommunications-Based Services. Environment and Planning B, 8, Raijas, A. (00). Te Consumer Benefits and Problems in te Electronic Grocery Store. Journal of Retailing and Consumer Services, 9, Hjortol, R. J. (00). Te Relation between Daily Travel and Use of te Home Computer. Transportation Researc Part A, 36, Zmud, J. & Arce, C. (July 000). E-Mobility (Electronic Travel): Measuring New and Evolving Beaviour. In Proceedings of te 9 t Conference of te International Association for Travel Beaviour Researc. Australia, Queensland. 30. Dijst, M., Farag, S. & Scwanen, T. (008). A comparative study of attitude teory and oter teoretical models for understanding travel beaviour. Environment and Planning A, 40, Farag, S., Scwanen, T. & Dijst, M. (007). Sopping online and/or in-store? A structural equation model of te relationsips between e-sopping and in-store sopping. Transportation Researc Part A, 4, Van, U. P. & Senior, M. (000). Te contribution of mixed land uses to sustainable travel in cities. In K. Williams, E. Burton and M. Jenks (Eds.), Acieving Sustainable Urban Form (pp ). London: E & F N Spon. 56

61 Transport and Telecommunication Vol. 4, No, 03 Transport and Telecommunication, 03, Volume 4, No, Transport and Telecommunication Institute, Lomonosova, Riga, LV-09, Latvia DOI 0.478/ttj SOME APPROACHES TO THE ROLLING WHEELS DYNAMICS MODELLING IN THE WEIGHT-IN-MOTION PROBLEM Alexander Grakovski, Yuri Krasnitski, Igor Kabaskin, Yuri Sikerzitski Victor Truacov Transport and Telecommunication Institute Lomonosova, Riga, LV 09, Latvia P.: (+37) Fax: (+37) Some possibilities of fibre-optic sensors (FOS) application for measuring te weigt of moving veicles realized in weigtin-motion (WIM) systems are discussed. As te first, te model of small-buried seismic sensor transient response excited by a car tyre interaction wit aspalt-concrete road pavement is proposed. It is supposed tat a seismic wave received by te sensor is te vertical component of surface Raleig wave. Te model is based on supposition tat a tyre footprint is acceptable to consider as some array of point sources of tese waves. Te proper algoritms permit to vary different parameters of te array excitation, as to footprint dimensions, load distribution, car velocities and oters. Te set of Matlab codes is worked out for seismic pulses modelling and processing. Te second way considered is to simulate te FOS signal in te basis of differential equations describing a deformable weel beaviour, or weel oscillations, in order to identify relations wit optoelectronic mecanical parameters. An attempt to find te mass of te veicle is based on minimizing te discrepancy between te actual FOS signal and te solution of te differential equation. Te accuracy of te evaluated weigt depends on many external factors, te matematical modelling of tem are expressed in te numerical values of te coefficients and external stimuli. Te influence of tese factors are analysed and tested by simulations and field experiments. One of ideas in dynamic weiging problem solution sould consist in evaluation of position of virtual gravity centre of te veicle in time. Te processing algoritm of te data received from te FOS is proposed based on conception of database retaining in some reference system memory. Certain requirements concerning te elements and blocks of te algoritm are defined as well. Te reference system is realized as te digital filter wit te finite impulse response. Te metod to estimate te filter coefficients is worked out. Several experiments wit tis algoritm ave been carried out for te veicle identification wit te reference loads adopted from real data. Te different factors ave an influence on te measurement accuracy of FOS. Te roadbed features, temperature, nonlinearities and delay effects in FOS are among tem. Te results of laboratory and field measurements wit FOS responses to different axle s loadings are presented. Carging and inertial caracteristics of FOS under te impact of various external factors (protective cover, temperature, contact area, and installation mode especially) as well as teir approximations are investigated. It is found tat te final calibration of te FOS as to be done individually and only after it as been installed in te pavement. Certain metods and algoritms of linearization, as well temperature and dynamic errors compensation of FOS data are discussed. Keywords: Weigt-in-motion, WIM systems, fibre optic sensors, sensor s sensitivity, tyre footprint, road pavement reaction, sensor response modelling. Introduction Economic situations in majority countries are visibly affected by range of automotive load transportation, and, terefore, by satisfactory condition of road networks. Degree of road deterioration is determined mainly by traffic activity and excessive brutto weigt of laden trucks. In accordance wit present standards, certain loading for a single axle of a motor car as to be restricted to some maximum allowable value. Exceeding of it assumes proper sanctions. Sensor networks for weiging road veicles are te tools for te problems of maintaining te quality of road surfaces solution in conditions of growt of traffic intensity in te European Union countries and creation of te traffic control intelligent transport systems. It is supposed tat te modern WIM system sould to be ensuring te measurement errors of 5% from true axle loading on te truck speed of km/ regardless of te sensor type. Very promising way to estimate te loading on certain axle is application of WIM systems [ 6]. Tere are many versions of WIM. Some from tem ave used sub-surfaced small-buried seismic sensors witin road pavement [3, 4]. Nowadays, te use of fibre-optic sensors, based on canging of te parameters of te optical signal from optical fibre strain under te weigt of passing transport [], takes an exclusive popularity in te problems of transport telematics. Tese sensors are relatively durable and inexpensive to manufacturing and operating. However, due to te low accuracy of measurement of weigt (especially in a moving object case), and ig dependence on weater conditions, tey are used mainly as a veicle s motion detectors only. 57

62 Transport and Telecommunication Vol. 4, No, 03 Output response of tat sort of sensor is a time series of some non-stationary transient signals. Te number of pulses is equal to te number of axles of te motor car wic is passed over te sensor at tat time interval. At first sigt all te pulses appear to be similar one anoter. However, some teoretical considerations, as well analysis of experimental data, demonstrate certain differences existing among tem. For successful designing a WIM system te nature of tese delicate differences ougt to be discovered and interpreted. Tus, te goal of a WIM system modelling is to work out a metod to filter individual differences in features of tese pulses. It is natural tat problems related to internal/external factors influenced on FOS transient forms ougt to be studied too. Tat is te range of questions supported in tis paper. Te organization of te paper is as follows. Some qualitative considerations concerning weelroad interaction and teoretical modelling of some possible FOS input forms are studied in Section. Te problem relates to calculating of a basic seismic pulse excited by a unit point mass body moving wit some friction along a orizontal road. Te model of tat pulse excited by tyre-road contact footprint is suggested. Te tyre footprint is considered to be some discrete dynamic array of surface Raleig waves sources wit exciting loadings distributed all along te footprint. If a sensor dept position (tat is a distance of it from pavement air surface) and tyre footprint dimensions are assigned as some initial conditions te proposed model permits to a) vary te function wic describes tyre contact pressure distribution inside te footprint contour, b) specify an automobile velocity, c) matc a rolling road resistance coefficient, d) take into account of wind velocity component oriented along te road, etc. Results of modelling are sensor response forms wic ougt to comply wit conditions specified above. Tese forms may be used as a starting material for formulation of te target inverse problem namely estimation of loads on individual axles of a veicle passed over te WIM sensor. Some potentials of tat model are demonstrated wit examples. Te aim of te Section 3 is in describing te approac to simulating te signal of FOS based on differential equations reflected deformable weel caracteristics [ 6] and identification of relations wit optoelectronic and mecanical parameters of a sensor. Finding te mass of te veicle is decided by minimizing certain discrepancy between te actual FOS signal and te solution of te differential equation of weel oscillations named above. Matematical modelling of external perturbed factors [6] is expressed in te numerical values of te coefficients and external forcing stimuli. Te impact of tese factors on te accuracy of WIM systems were analyzed and tested by simulations and field experiments. Te Section 4 is devoted to study an adaptive metod for system identification in order to determine te type and weigt of te veicle. Tis metod consists in using a database of signals obtained wit te fibre optic sensor wen a truck of te certain weigt driving on it. Signal data from te output of fibre optic sensor based on comparison of te distribution center of mass along te axes of te veicle in te static and dynamics are reduced to a single scale on te time axis. Tis procedure uses data on veicle speed. In its turn, comparison of signal output from te fibre optic sensor for dynamic and static weiging using adaptive filtering tecniques, allows us to create a database of equivalent reference systems in te form of a set of stationary filters for trucks wit te certain weigt. After processing te signals using te proposed metod by comparing te response at te output of a stationary filter, simulating te operation of te master system and te response to te output fibre-optic sensor, after te passage of te veicle of te unknown type, weigt of te truck is estimated. In Section 5 extensive studies is carried out of discrepancies in pulse forms registries from FOS as te results of laboratory and field natural experiments under different initial conditions (i.e., truck loads and velocities, as te weater too). Te results of te laboratory and field measurements are presented, as well as te load and inertial caracteristics and teir approximations, obtained for fibre-optic sensor under te impact of various external factors (protective cover, temperature, contact area, and especially installation). We discuss te algoritms for linearization, calibration and metods of temperature compensation of fibre-optic sensors wen using tem in weigt in motion (WIM) applications. Last trends on WIM issues indicate tat FOS sensors are based on two main principles: Bragg grating (te cange of diffraction in a cannel under deformations) [7] and cange under deformations of te fibre optical properties (transparency, frequency, pase, polarization) [7]. Te cange of transparency (te intensity of te ligt signal) of SENSORLINE experimental sensors [8], as basic operating principle, is considered in tis study. Most popular are te multi-sensor systems (MS-WIM) [9], were te accuracy of measurements is obtained by statistical processing of 6 0 or more sensors, situated in sequence into te pavement at some distance to eac oter. Obtaining te measurement errors of < 5% tis approac takes te expenses, 58

63 Transport and Telecommunication Vol. 4, No, 03 compatible wit te bending plates and needs te reconstruction of roadside surface. Te aim of te present study is to obtain te measurement errors of 0% witout multiplying of sensors. Tis work was worked out by supporting of ERDF funding project Fibre Optic Sensor Applications for Automatic Measurement of te Weigt of Veicles in Motion: Researc and Development (00 0).. Transient Response of a Small-Buried Seismic Sensor Output response of tat sort of sensor is a time series of some non-stationary transient signals. Te number of pulses is equal to te number of axles of te motor car wic is passed over te sensor at tat time interval. At first sigt all te pulses appear to be similar one anoter. However, analysis of experimental data, demonstrate certain differences existing among tem. For successful designing a WIM system te nature of tese delicate differences ougt to be discovered and interpreted. Te goal of a WIM system designing is to work out a metod to filter individual differences in features of tat pulses. In tis section te algoritm is developed based on modelling of a small-buried seismic sensor response excited by forced impact distributed along a car tyre footprint to aspalt-concrete road pavement. It is supposed tat a seismic wave perceived by te sensor is te vertical component of surface Raleig wave [7, 8] propagating in te road pavement top layer. It is proposed te tyre footprint sould to be consider as some discrete array of point sources of surface Raleig waves eac wit own exciting loading distributed along te footprint. A transient signal created by interferences of tese waves is received by te seismic sensor... Some Features of Weel Road Interaction A qualitative picture of weel road interaction (Figure ) is set out in [9]. Tis interaction takes place witin te bounds of some contact pavement surface referred to as a tire footprint. An axle loading W is translated to a footprint as a non-uniformly distributed weel pressure. It is described by a certain function F p (x, y), wic is depended on tire elasticity forces depicted by even temporal modes, as well as friction forces caused by elements of a pneumatic tire tread. As te weel is in rolling motion along te axis x wit some linear velocity V and angular velocity ω w, te radial deformation of te tire tread is canged abruptly in time wit a really complicated manner. It is amplified in te tread front part, but it is reduced in te back part of it. Tese reasons are reflected in te friction forces as te dominated odd modes. As a result, at te a-c part of footprint te friction forces ave te same directions as te tire elasticity forces, but along te c-b part teir directions are opposed. Terefore, weel pressure F p and road reaction R z diagrams are non-symmetrical relative to te vertical axe z passed troug te weel centre. Te extreme of pressure and reaction forces are displaced from te axe z by te sift s. Tus, a value of s sould serve as a measure of weel road interaction. Figure. Weel road interaction picture (adapted from [9]).. Response on a Unit Point Mass Movement Unfortunately, exclusive multifactor complexity of considered process not yet permits to create te matematically strict description of it. Te problem described in tis part is derivation of a seismic pulse form excited by a unit point mass body moving wit certain friction along a smoot orizontal road. Figure sows te sketc of te task. An omni-directional (isotropic) seismic sensor is placed on some dept from road surface and it is superposed wit te origin O of Cartesian coordinates system. A point body moves from initial position t = 0, x = x 0 to rigt along te plain pavement surface wit constant velocity V. It is supposed tat x 0 is a negative value. Te body experiences an influence bot force of weigt W and friction force F too. 59

64 Transport and Telecommunication Vol. 4, No, 03 x Instantaneous position of te body concerning te sensor describes by distances x, r and an angle α were / = x 0 + Vt, r = ( x + ), sin α = / r, cos α = x / r. () Movement of tis body excites seismic oscillations in pavement layer propagating along te road surface as te Raleig type waves wit velocity V R. Te current pressure P perceived by te sensor depends on te instantaneous sum of projections onto running radius vector r of te forces W and F. It is seen from Figure tat projections of W would be alter teir directions as soon as te sign of x is canged, ence, one can describe P = ( W sin α F cos α ) / r, () were it as been taken into account te reverse squared root dependence of Raleig surface wave intensity from distance [4, 5]. If a sensor as to response on normal, or z, component of force P only, ten it is necessary to project te force in Eq. () to z-axis. Allowed for (), te result can be written as P + 5/ z = W ( kf x / ) / r, (3) were te value of k F = F/P may be considered as a rolling friction coefficient. As te Raleig surface wave propagates along te r from te point of instantaneous position of te moving body to te sensor it sould be delayed in time on t R = r/v R. Temporal scale of sensor as to take it into account. Terefore, sensor time t s as to look as t s 0 0 Figure. Excitation of seismic sensor by a unit point mass body movement = ( x x ) / V + ( r r ) /, (4) V R were r 0 = (x 0 + ) /. Equations (3) and (4) may serve as a basis for modelling of sensor responses time forms initiated by a moving body. Certain results are presented on Figure 3 wit some variations of k F. It sould be noted tat in accordance wit Eqs. () and (4) te distance x and sensor time t s are in nonlinear relation due to te second item in (4). It is especially significant in a region of small times. Figure 3. Pressures pulses forms versus rolling friction coefficients 60

65 Transport and Telecommunication Vol. 4, No, 03 Te plots on Figure 3 represent te results of seismic wave propagation in aspalt-concrete road pavement. Te exact value of Raleig wave velocity is unknown but it peraps is about V R = 400 m/s [0]. It is seen from te left part of Figure 3 tat te growt of friction coefficient leads to certain asymmetry of sensor transient response wit respect to maximum value of it. It is accompanied by increasing of pulse amplitudes, durations and areas. In te same time te maximum of te pulse is sifted wit some lag in opposition to direction of motion. It is not contrary to [9]. Some features of te pulses are sown on graps placed in rigt panel of Figure Seismic Pulse Excited by Tyre-Road Contact Footprint Te model described in te previous part as been permitted to derive a seismic pulse form excited by a unit point mass. Essentially, solution of tis task sould be considered as a certain Green s function. Hence, it may be used to find a response caused by motion of some finite-dimensional body wit known distribution of mass along it. Suc te solution is reduced to modelling of interaction of seismic pulses excited by different parts of te moving body taking into account te lags. Tat approac may be conformable wit te problem of sensor response excited by forced impact of a tyre to pavement. Unfortunately, any analytics, neiter of mass distribution along a footprint, nor road reaction, are not discovered in accessible reference sources. Te normalized function sin( πx / x W ( x ) = cos[ π( x x max max ), ) / ( l x max )], x [ 0, x x [ x max is proposed ere to descript te distribution above as a piecewise smoot approximation were l is te footprint lengt, x is te current coordinate along footprint. Te value of x max is te position of maximum pressure point, or road reaction (see te sift s on Fig. ). Tis maximum is displaced from te point of footprint centre in te direction of motion of te car. Te value of s would be associated wit friction coefficient and depends on road rolling resistance, car velocity, aero dynamical factors, wind vector etc. However, te numerical modeling practice ave been demonstrated tat more promised results ave to take place wit upgraded formulation of (5), namely α W M ( x ) = W ( x ), α <. (6) More pertinent values of α about ave been establised wit numerical experiments. Beavior of (6) under different α is illustrated in Figure 4. It is conformed to friction coefficient value nearly 0.3. Te curves on Figure 4 are not contrary to proper graps wic ad been given in some reference sources (see Fig. ) to explain nature of road reaction on a footprint contact at least in qualitative sense. In order to calculate sensor transient response initiated by a body wit finite sizes it is necessary to consider an equivalent seismic source in te form of - or -D discrete array. It consists of well-defined number of point sources eac wit own exciting loading. Every element of te array excites own Raleig wave and makes some contribution into te normal pressure component influenced on a sensor. max, ] l ] (5) Figure 4. Normal road reaction distribution (l = 0.3 m, x max = 0. m) Figure 5. Pulse response from tyre footprint contact wit a road 6

66 Transport and Telecommunication Vol. 4, No, 03 Described approac permits to find a sensor response on sort-term impact under simultaneous contact of tyre wit a pavement inside footprint. Te rigt-angled contour of te latter is supposed. Symmetric load distribution along te footprint is sown on Figure 5. Te middle panel presents relative contribution into total pressure from different elements of an equivalent array. It is supposed te sensor is situated on te dept = 0.05 m from te surface. Te simplest case as a place tat te footprint widt is supposed to be a constant. Coordinates along footprint are counted from sensor position (as te value of x on Fig. ). Finding of sensor reaction in tose conditions is, in essence, a one-dimensional problem consisting in evaluating of te pulse response of te sensor wit delta stimulus. Suc te response after normalization is plotted on Figure 5, te lower panel. It would be noted tat natural experiment by means of te direct instrumental measurement of te pulse response is rater difficult in realization. It is possible tat computer modeling is only one te way to decide tis problem by relatively inexpensive tools..4. Variations of Seismic Pulse Forms As a function of weigt, a seismic pulse form in WIM applications as to depend on many arguments and can be described by different features. Being realized as computer codes te models considered above allow analyzing relative influence of some factors on inerent structure of te pulse and picking out te most correlated wit WIM aims. In computational sense, finding of a pulse form excited by a moving array wit N elements reduces to calculation of some matrix. Every string in it is an elementary pulse. It is conformed to appropriately delayed motion of an individual element of te array (see Eqs. ( 4)). Tat matrix is P P K P P P 0 = K K K 0 K 0 P P K K K N P 0 K K K 0 0 K K P K K K N,K 0 0, K PNK (7) were K is te prescribed number of array element positions, or sifts, along x-axis on Figure. Every string in (7) is completed by zeroes in start and finis ends in order to equalize te lengts. Current position of te element P NK determines te total lengt of any string. As sensor time t s depends on x, nonlinearly plain addition by columns in (7) is not correct to ave te rigt response form. Certain interpolation for every string as to be done preliminarily using te Raleig wave minimal arrival time /V R as te step. It can be observed tat computation of te matrix (7) is equivalent to convolution P( x ) = P z ( x ξ )W M ( ξ ) dξ (8) of te unit point pulse (3) wit te footprint loading distribution (6). Te latter sould be considering as a function of time wic is revealed oneself in process of footprint rolling over te sensor. As a sounding example, tis metod may be applied to te problem of correlation of pulse form wit footprint lengt solved by computational experiments. In order to correspond wit (7) it as been supposed te seismic signal is formed by linear combination of some delayed pulses. Eac from tem is excited by proper element of discrete array wic is considered as a certain equivalent of te footprint. Te veicle velocity V, as well as te position, or sift s from a loading centre, of te road reaction R z maximum value (see Fig. ), are treated as te constants. Results are sown in Figure 6. 6

67 Transport and Telecommunication Vol. 4, No, 03 Figure 6. Forms of pulses vs footprint lengts In particular, it is seen from te rigt panel of Figure 6 tat durations of te pulses and displacements of te pulse maxima depend on footprint lengt rater linearly, in contrast to oter caracteristics of tese transients. Te oter initial conditions to modelling were used to draw te graps on Figure 7. It is seen tat variation of maximal loading sift in (6) as a predominant influence on te pulse maxima displacements only. Te rest of parameters are canging just insignificantly. Figure 7. Forms of pulses vs maxima force sift 63

68 Transport and Telecommunication Vol. 4, No, 03 Dependencies of te modelled seismic pulse parameters from car velocity are sown on Figure 8. One can observe some weak linear relation among velocity and maximal values of te pulses. Te oter parameters ave noticeable nonlinearities wit respect to tis velocity. Figure 8. Forms of pulses vs car velocity In common case, if a sensor dept position and tyre footprint contour dimensions are assigned as some initial conditions te proposed model permits to: a) vary te function wic describes tyre contact pressure distribution inside te footprint contour, b) specify an automobile velocity, c) matc a rolling road resistance coefficient, d) take into account of wind velocity component oriented along te road, etc. Te facts and regularities ascertained by modelling and described above may be elpful in designing of WIM systems. In particular, it may be used as a starting material for formulation of te target inverse problem namely estimation of loads on individual axles of a motor veicle passed over te WIM sensor. 3. Response of Seismic Sensor Signal Modelled by te nd Order Electrical Circuit From te teory of deformable weel bearings [ 6] tat te weel side of te road is normal and tangential distributed reaction. Complete description of te dynamics of te mass of te veicle forms a unified system of differential equations tat take into account te interaction of two masses rotating and moving steadily. Te total number of second order differential equations is six, two for eac spatial coordinate. Into account for te problem of determining te mass car all times, te drag coefficient of te road, rolling friction, etc. unreal. Simplifying te equations of motion of a deformable weel go to te single-mass oscillation system tat takes into account only te motion of te unsprung masses in te vertical plane. Assuming tat te motion is on te way to te armonic micro profile, we obtain [4]: d z dt dz α + ω n z = ω n ε tire q cosυ t. (9) dt

69 Transport and Telecommunication Vol. 4, No, 03 Here z is te coordinate of vertical movement of te masses, α = is coefficient of damping of m te unsprung masses, m is unsprung weigt of te veicle, ωn = ( Cbr + Ctire ) / m is natural frequency of te unsprung masses, C br is suspension s stiffness, and C tire is stiffness of te tire, ε tire = C tire / ( C tire + C br ) is smooting factor of te tires ( ε tire 0,6), q 0 is maximum eigt of bumps in te road, and finally, ν is te frequency of forced armonic tremors from te road surface. Try to solve te problem of simulation of tis equation in te form of a circuit wit an unknown input to caracterize te effect of mass, friction, stiffness of tires and suspension, etc. Te first attempt to determine te mass of te veicle and te frictional forces acting on te weel is a WIM-signal approximation in te basis of certain matematical functions. Signal approximation by polynomials of various degrees and identify teir coefficients depending on te weigt and friction gives not a positive result. In te armonic approximation basis also found no linear relationsip between te weigt of te car and te parameters of te armonic series, owever, possible to determine te own te frequency of te car on te WIM-signal, wic eventually led to te idea to try to solve te problem directly in te basis of differential equations. In te first pase was implemented simulation WIM-signal wit RLC-circuit of te second order (Fig. 9) wit external action in te form of a trapeze wit controlled quantities of te front, fall, te flat top and amplitude. K a Figure 9. Simulation of WIM-signal as output reaction of te second order circuit Te figure sows part of te output response and te external action in te form of trapezoidal (Fig. 9). Since te output response is like to WIM-signal, we can try to find a relationsip between te mecanical parameters of differential equations of te teory of deformable rolling weels and its electric counterparts, including te weigt of te car wit te numerical values of te circuit elements. However, tere is muc more productive idea to direct solution of linear (and peraps non-linear) differential equations wit forcing action in trapezoidal (or oter sape). If te output variable of te equation to approximate te WIM-signal, optimising te parameters of te differential equation and find te relationsip between te mass of te veicle and te caracteristics of te driving influence, te problem will be solved. Among MATLAB program was written for solving differential equations wit forcing input action in te form of a trapezoid. Te objective function wit a weigted least square value and te maximum deviation between te WIM-signal and te output response of te differential equation was formed. A program of global to minimize te objective function was designed, since te attempts to use existing local metods in MATLAB not give te desired results. Tis approac was applied to te real signals obtained from measurements carried out by te project Fibre optic sensor applications for automatic measurement of te weigt of veicles in motion: researc and development (00 0). 3.. Data Processing Based on Optimisation of Modelling Input Pulse In a series of picture below (Figure 0), as an example, is sown te rigt track WIM-signal s4_70km_0_04_0 and te results of optimisation for eac weel trucks wit known weigt for axes in te static (in coordinates: seconds, volts). Te objective functions before and after optimisation f0 and f00, and te vector of optimal parameters x00, including te values of breakpoints for trapezium and coefficients of te differential equation of second order are printed too. 65

70 Transport and Telecommunication Vol. 4, No, 03 Figure 0. WIM-signal (rigt weel), optimal output reaction and trapezoid input actions Te results of calculations based on global optimisation algoritm are as follows: st axis. f0 =.945e+00, f00 = 3.854e-00, x00 =.339e+000, st axis. f0 = 5.33e+00, f00 =.89e+000, x00 =.867e+000, 3 st axis. f0 =.6977e+00, f00 = 3.86e-00, x00 =.3577e+000, 4 st axis. f0 =.5870e+00, f00 = 4.393e-00, x00 =.550e+000, 5 st axis f0 =.733e+00, f00 = 6.949e-00, x00 =.966e+000. Initial approximation, te vector x0 = {t, t, t3, u, d, d} wit te values of te front, flat top and fall trapeze t = 0.5, t = 3, t3 =, representing te eigt of te trapezoid u = U max and te coefficients of te differential equation d =, d =, (in (9) α = d + d, ω = d*d)), is te same n for all axes. Note tat te actual parameter d = 0.0 * d. 3.. Calculation of te Axle Mass Based on Results of Optimisation Te next task is to process te data using different speeds of veicles wit known mass and finds te correspondence between te parameters of trapezoidal signal, te coefficients of te differential equation and te mecanical parameters mass, friction force, stiffness of tires and bracket, etc. Known mass to te rigt track of te axes in te static are m = {3680, 6380, 980, 830, 955} (kg). We use te known mass of te st axis to determine te unknown values in (9): 66

71 Transport and Telecommunication Vol. 4, No, 03 Ka = (d+ d) m, (0) ωn = d d = ( C br + C tire ) / m. () From (d =.039/00; d =.948; m = 3680 kg) we obtain Ka = (d + d)*m = 73.9kg, and C bt = C br + C tire = m ( + d d) = 3765.kg. We apply tese factors (suc as calibration) to determine te mass of -axis for te signal s_50km_0_04_0 (d = ; d =.305). Now we obtain m = Ka / (d + d), m = e m = Cbt / ( + d * d), m = 3.643e (S - m = 368kg). Te coefficients K a and C bt for 5 t axis (d = ; d =.583; m = 830 kg) cfn be calculated as: K a = (d + d)*m = e + 003, C bt = ( + d*d)*m =.8445e We apply tese factors to determine te mass of te 5 t axis signals: s_50km_0_04_0 (d = ; d =.8494) m =.465e + 003, m =.803e+003. (S - m = 880 kg) s_70km_0_04_0 (d = ; d =.5898) m =.837e + 003, m =.8440e (S - m = 880 kg) s3_90km_0_04_0 (d = ; d =.66) m = 3.739e + 003, m =.896e (S3 - m=965 kg) It is seen tat te obtained value is closer to te mass m and is located to an accuracy of less tan 5%. Let s verify tese findings for te tird axis, taking te calibration signal: s4_70km_0_04_0 (d = e-004; d =.493e + 000; m = 980 kg). Ten K a = (d + d) * m = kg; C bt = ( + d * d) * m = 98 kg and te mass of te tird axis for te various signals is. s_50km_0_04_0, (d= 4.036e-003; d =.7978) m =.467e+003, m =.9605e+003; (m = 3045kg), relative error = e-00,. s3_50km_0_04_0, (d=.895e-003; d =.7360) m =.637e+003, m =.9666e+003; (m = 900kg), relative error = -.966e-00, 3. s3_90km_0_04_0, (d =.673e-00; d =.5788) m =.76e+003, m =.8949e+003; (m = 900kg), relative error = e-00, 4. s_70km_0_04_0, (d = e-003; d =.67) m =.6475e+003, m =.948e+003; (m = 945 kg), relative error = e-004, 5. s5_70km_0_04_0, (d = 7.044e-003; d =.9333) m =.90e+003, m =.946e+003; (m = 967,5 kg), relative error = -.545e-003. Figure. Elastic caracteristics of te tire air pressure is 3 kg/cm (ere F cw is an elastic force in tires), and loading, unloading curves [4] 67

72 Transport and Telecommunication Vol. 4, No, 03 So, as it was seen, in all of calculations, te relative error wasn t exceeding 3% of measured weigt s value but it isn t te limit. For example, it is a grap (Figure ) for te ZIL-30 truck veicle s tires, noting tat te damping capacity of te tire due to intermolecular friction rubber and internal friction between te elements of te tire [4]. Te presence of friction leads to te ysteresis loop on te elastic caracteristics of te tire, wic becomes narrower wit increasing velocity. Te knowledge of static stiffness values for te currently operating commercial veicles would be elpful for implementation of te results of tis approac or its modified version into weigt-in-motion (WIM) tecnologies because of ig accuracy (till 3%) of it. 4. Problem of Decision Making Concerning te Overloaded Veicle Selection 4.. Formalization of te Decision Making Problem One of te problems of te dynamic weiging is te task of te selection of te overloaded veicles in te traffic flow. In tis case, te weiging task will be transformed into te decision making task if te veicle is overloaded or not. Te decision-making problems are usually solved by comparing te current event wit some te so-called reference one. In te systems wit electronic sensors, te standard of reference can be specified eiter in te form of te reference signal, or as a device, wic carries te caracteristics of te reference signal. In te dynamic weiging systems were fibre optical sensors are used, te reference signal from te sensors output sould ave te caracteristics tat correspond to te true weigt of te veicle detected by te sensor. As tere are random and regular destabilizing factors tat distort te signal from te fibre optical sensors output, te reference signal can only be formed as te result of te averaging signals from te output of te same sensor due to repeated overriding of te veicle loaded wit te reference weigt. Tus, te reference signal tis is te response of te sensor in some measuring system, tat corresponds to te averaged response of te sensor of te real system to te repeated overriding of te veicle loaded wit te reference weigt. Te corresponding conditional system will be called te reference system. After te reference system is created, te decision making tasks concerning te veicles accordance to te allowable weigt for a given road can be made, for example, by te correlation analysis tecnique [5]. Tis calls to set te model of te reference system and to set its parameters in suc a way, tat te output signal would ave caracteristics tat corresponds to te maximum of te allowable weigt of te veicle for te particular type of te road. Comparing te responses from te sensors output of te real system, wen te sensor detects te passing veicle only once, wit te response at te reference systems output, using te correlation metod, it is possible to determine wat event rate is bigger as te veicle te greater weigt tan existed in te reference system or te smaller one. 4.. Reference Dynamic Weiging System and its Features As stated above, te decision making task concerning te veicles overloading can be solved by comparing te sensors response of te real systems output wit te sensors response of te reference system. Besides, te sensors response of te reference system sould be in accordance wit te average response of te real systems sensor to te repeated overriding of te veicle loaded wit te reference weigt. Te creation of te adequate reference system is a complex tecnical callenge tat requires taking into account all te destabilizing factors tat mentioned above, wic affect te WIM sensors. And, as among te destabilizing factors tere are veicle specifications tat lead to te errors in te sensors readings, te design of te reference system can only be applied to te specific and pre-selected veicle Figure. Te response of sensors output, wen te empty truck passes it at te speed of 90 km/ 68

73 Transport and Telecommunication Vol. 4, No, 03 From te figure, sowed above, it is possible to see, tat wen te truck passes troug te sensor placed into te road, eac of te five axles of te veicle appears at te sensors output as leap of te voltage (vertical axis). Te set of five leaps of te voltage on te time axis (te orizontal axis sows te numbers of discrete signal samples in time) is te system response from te passage of te selected veicle type. It is also can be seen tat te responses of te sensor wile passing te empty veicle and te loaded veicle, wit te reference weigt, are different. However, te degree of te responses similarity or difference can be determined statistically only. Te procedure of te comparison of responses is complicated by te fact tat te speed of te veicle wen driving over a sensor is not known in advance and, terefore, te location of te voltage leaps on te time axis is unknown. Tese circumstances brings us to te need of te data pre-processing in te fibre optical sensor, for te purpose of bringing te sensors output responses into te one and te same time frame. As well, in te same time frame sould be generated te reference systems response from te veicle of te given type, tat is loaded wit te reference weigt and tat passes te sensor. Te device tat generates tis reference response is te reference system. To bring te response from te sensors output into te same time frame, it is needed to make te following calculations: Calculation of te centre of mass of statically weiged veicles of a given type. Calculation of te centre of mass of dynamically weiged veicles of a given type. Calculation of te speed of te veicle of a given type. Basing on te assessment of te decentralization of te mass of te dynamically weiged veicle wit respect to te centre of mass of te statically weiged veicle, and, knowing te speed of te latter, te peak voltage amplitude response is recalculated from te sensors output in te new coordinate system, wic corresponds to te single time frame. To reduce te errors associated wit te temperature sensor instability, during te recalculation of te peak voltage amplitude response from te sensors output, tere are calculated and introduced te correction coefficients. Te transformation procedure to te same time frame of te WIM sensors response is sufficiently completely described in [6] Reference Signal Normalization and te Conception of te Reference System As te reference signal it sould be te response from te WIM sensors output tat is normalized to te common time frame, and contains te information about te veicles weigt tat passes te sensor. Te reference signal is generated by te averaging te values of te voltage peaks amplitudes response from te sensors output resulting from te repeated calibrating overriding of te specified veicle type wit te reference load. In tis case, for te averaging procedure it is very important to know, ow many sensors are installed for being passed by te veicle, or ow many times te veicle passes te sensors, all of tis will result te averaging. Tus, tere is an opportunity to get two signals in one time frame. Te first signal corresponds to te distribution of te voltage peaks amplitudes of te statically weigted veicle, and te second one to te distribution of te voltage peaks amplitudes in te reference signal. Te work of te WIM, in tis case, can be interpreted as te work of some equivalent electronic system. At te input of tis system was sent te reference signal and te response of tis system is te signal tat corresponds to te distribution of te voltage peaks amplitudes of te statically weigted veicle of te specified type and wit te reference load. In wat follows, tis system will be called te reference system. Main element of te reference system is te filter wit te finite impulse response, wic transforms te reference signal into te corresponding distribution of te peaks amplitudes of te statically weigted veicle of te specified type and wit te reference load. Te quality (accuracy) of te transformation depends on te correctness of te definition of te filters weigt coefficients. Due to te complexity of te exact analytical description of te initial signals it was offered to receive te adequate value of te filters weigt coefficients by te metod of systems identification [5]. On Figure 3 te structure of te algoritm is sown for defining te caracteristics of te reference system. Main element of te structure is te filter wit variable weigt coefficients (te transversal filter). On te filters input te reference signal is sent. Te task of te algoritm is following: to minimize te error of te reference signal reproduction by adjusting te parameters of te transversal filter. As te reference signal, te signal was taken, tat corresponds to te distribution of te voltage peaks amplitudes of te statically weiged veicle of te specified type and wit te reference load. 69

74 Transport and Telecommunication Vol. 4, No, 03 To obtain te weigt coefficients of te transversal filter were used te least squares metod [5]. Tis adaptive algoritm was taken due to te fact tat it works in real time, i.e., tere is no need to accumulate errors from te sensors. Te use of oter adaptive algoritms, suc as normalized LMS, variable step size LMS and sign algoritm LMS, do not lead to te significant improvements in te calculation of weigt coefficients. Figure 3. Te structure of te adaptive algoritm for determining te weigt coefficients of te transversal filter Calculation of te coefficients is carried out according to te formula () w( k + ) = w( k) + μ e( k) u( k), () were µ is positive coefficient, wic is called te size of te step, e(k) is te replication error of te reference signal, u(k) is te content vector of te delay-line on te k-step. Later, tese weigt coefficients are stored in te form of te constant filters weigt coefficients of te reference system Overloaded Veicles Identification Results During te test of te proposed algoritm tere were formed two reference systems (for te reference signal of te veicle tat is fully loaded; and for te reference signal of te empty veicle. Table. Veicle passing options wit different weigts Number of empty veicle passing Number of loaded veicle passing Discrete filter Nr. (reference of empty veicle) Speed, 70km/ Speed, 90 km/ Discrete filter Nr. (reference of loaded veicle) Speed, 70 km/ Speed, 90km/

75 Transport and Telecommunication Vol. 4, No, 03 Te converted signal from te fibre optical sensors output was sent to te discrete filters of te corresponding reference systems, and tere were estimated te relative standard mean square errors on te filters outputs. In te Table, tere are sown te following results tat were received during te working process of te algoritm of te systems identification. As one can see, not all te results are well-defined. Tis could appen due to te fact tat te fibre optical sensor was not tested for te spatial omogeneity of te sensor response over te period of te working zone. In te data used, te possible errors can occur. Tis is associated wit te fact tat te signals were not recorded in one day, but for a certain period of time, wen te connection and te disconnection of te equipment from te sensors were placed. According to te results, it is possible to conclude tat te designed system copes wit te task of te veicles identification by te excess of te permitted maximum load on te road. 5. Weigting-In-Motion of te Axles of Veicle by te Area Metod 5.. WIM Basic Caracteristics Te fibre optic force sensor is a cable consisting of a potoconductive polymer fibres coated wit a tin ligt-reflective layer (Fig. 4). A ligt conductor is created in tis way, from wic te ligt cannot escape. If you direct a beam of ligt to one end of te cable, it will come out from te oter end, and in tis case te cable can be twisted in any manner. In order to measure te force acting on te cable, te amplitude tecnology is more appropriated for measurement, wen measures te optical pat intensity, wic canges wile pusing on te ligt conductor along its points. At tese points te deflection of a ligt conductor and reflective coating occurs, tat is wy te conditions of ligt reflection inside are canged, and some of it escapes. Te greater te load te less ligt comes from te second end of te ligt conductor. Terefore te sensor as te unusual caracteristic for personnel, familiar wit strain gauges: te greater te load te lower te output is. Apart from te fact tat it is reversed and in addition to tis it is non-linear. Let to avoid te inaccuracy of zero load level we need to exclude te ig frequency components from te voltage signal at te output of te sensor s transducer by filtering as well as to recalculate te voltage signal U(t) (Fig. 4(a)) into te relative visibility losses signal V(t) (Fig. 4(b)), directly related to te weigt pressure on te FOS surface, by te transformation (3): V ( t ) U U( t ) 0 =, (3) U 0 were U 0 is te voltage of sensor output wit zero load. Te signal transformation to te relative visibility losses signal V(t) gives te possibility to compare signals for different measurements in different conditions. Fibre optic load-measuring cables are placed in gap across te road and are filled wit resilient rubber (Fig. 4).Te gap widt is 30 mm. Since te sensor widt is smaller tan te tyre footprint on te surface, te sensor takes only part of te weigt axis. Two metods are used in te existing systems to calculate te total weigt of te axle [7, 7]: te Basic Metod and te Area Metod. Te following formula is used to calculate te total weigt of te axis using te Basic Metod: W a = At Pt, were W a is te weigt on alf-axle, A t is te area of te tyre footprint, and P t ~ V(t) is te air pressure inside te tyre and, according to Newton s 3 rd law, it is proportional to te axle weigt. As we can see te exact values of te formula factors are unknown. Te area of te tyre footprint is calculated rougly by te lengt of te output voltage impulse, wic, in its turn, depends on te veicle speed. Te Area metod uses te assumption tat te area under te recorded impulse curve line, in oter words te integral, caracterizes te load on te axle. To calculate te integral, te curve line is approximated by te trapezoid. In tis case te smaller te integral te greater te load. Tis metod does not require knowing te tyre pressure, but it requires te time-consuming on-site calibration. Also, it as to be kept in mind tat te time of te tyre passing on te sensor is too small to get an electrical signal of ig quality for its furter matematical processing. We use te Area metod for footprint area definition in (4), but te pressure is measured from te signal amplitude. (4) 7

76 Transport and Telecommunication Vol. 4, No, Left Rigt v=60 km/ 7.4 (a) (b) Left Rigt v=60 km/ x (c) Left Rigt v=60 km/ Figure 4. Location and waveform of te SENSORLINEE PUR fibre optic sensor [8], (a) Voltage, V (b) Visibility losses, (c) sensor s s position against te weel and weel s footprint, (d) measured parameters for weigt calculation 5.. Field and Laboratory Experiments wit FOS and Weigting-In- -Motion Results Tere was te set of measurement experiments wit te roadside FOS sensors on April, 0. Loaded truck (Fig. 5) was preliminary weiged on te weigbridge wit te accuracyy < %. Figure 5. Experimental truck Volvo FH wit full loadd kg 7

77 Transport and Telecommunication Vol. 4, No, 03 Table. Etalon axle s weigt (tons): Date: (Air Temperature + о С) Etalon weigts of separate axles are given in te Table. Te signals from output of FOS sensors for truck speed 70 km/ and 90 km/ are demonstrated on Figure 5. It is evident tat te signals for different speed are strongly canged by amplitude and te proportion of amplitudes does not fit te axle s weigts (Fig. 6). Te reason of tis beaviour may be concluded in te FOS properties suc as weigt (pressure) distribution along te sensor s lengt as well as sensor s non-linearity and temperature dependence Left weel V = 70 km/ Rigt weel V = 90 km/ Left weel Rigt weel Figure 6. Examples of FOS signals of experimental truck for veicle s speed 70 km/ and 90km/ respectively Load caracteristic was measured from te SENSORLINE PUR fibre optic sensor by means of a SL Transducer (optical interface) optical signal analyser as been developed by SensorLine GmbH [8]. For obtaining te static load caracteristics of te sensor te MTS compression macine of te Latvian University Institute of Polymer Mecanics was used. Effort in te range kg was transmitted to te sensor troug a steel plate (see Figure 7 (a)) te size of 00 mm by 00 mm at te temperature conditions of +8 o C. Te area of application of te force to te sensor was 30 mm to 00 mm. Load curves (gradual increase of te force) and unloading (gradual decrease) of te sensor are sown on Figure 7 (b). Te curves clearly demonstrate te presence of nonlinear beaviour te so-called ysteresis loop, or te difference between te curves load-on and load-off. Te reason for tis is te residual deformation of rubber protective sensor ousing. Tese curves are conveniently approximated by yperbolic tangent function: V ( p ) = a0 + a tan( ω p + ϕ ), (5) were p is te pressure on te sensor s surface but a 0, a, ω and ϕ are te parameters of approximation. For example, load-on curve from te Figure 7(a) conditionally describes by te next approximation coefficients: a 0 = , a = , ω = 0.480, ϕ = , calculated by te metod of least square criteria optimisation. For te local approximation is quite suitable to use cubic or quadratic approximation polynomials, defined by te criterion of least squares. 73

78 Transport and Telecommunication Vol. 4, No, 03 For example, load-on curve fromm te Figure 7(a) conditionally describes by tee next approximation coefficients: a 0 = , a = , ω = 0.480, ϕ = , calculated by te metod of least square criteria optimisation. For te local approximation is quite suitable to use cubic or quadratic approximation polynomials, defined by te criterion of least squares Load Down (plate 0 cm) Load Up (plate 0 cm) Load Down (plate 40 cm) Load Up (plate 40 cm).8.7 Ratio "Up" FOS Visibility (a).3. Ratio "Down" Ratio Average (b) Pressure (kg/cm) Figure 7. (a) Experimental static load caracteristic of FOS wit doubling of active area and (b) Ratioo between two curves by te lengt of 00 and 400 mm respectively Anoter experiment was associated a wit te influence of sensor data, d producedd by te dual weels. Comparison of load curves at tee impact area troug te steel plate wit te size of 00 mm by 00 mm and 400 mm by 4000 mm is sown on Figure 7(a). Te impact area on te sensor is increased by times, and te losses of transparency of te FOS, ass can be calculated from te relative position of te curves (Figure 7(b)), decreases approximately by π/ π times, due to te radial distributionn of load along te contact surface of te steel plate and te sensor. (a) T +8 deg. Down T +8 deg. Up T -0 deg. Down T -0 deg. Up FOS Visibilty (b) Pressure, (kg/cm) Figure 8. (a) Sceme of temperature factor measurement and (b) Experimental static load caracteristic of FOS wit doubling of activee area of weel s footprint In actual conditions of use te fibre optic sensor (FOS) wit te protective p rubber ousing is i built into te road surface and, terefore, affected by temperature, wic canges te stiffness of rubber. r An experiment determining tee effect of temperature s factor is to obtain load caracteristics and comparison wit FOS at 8 C and after cooling in te freezer camera down to -0 o C. Te results of tese measurements are sown on Figure 8(b). 74

79 Transport and Telecommunication Vol. 4, No, 03 Due to canges in te cover ousing rubber stiffness, te measurements of pressure canges up to approximately 50% and increased residual effects (ysteresis). Consequently, it is necessary to use te temperature correction in te measuring sceme. Anoter source of measurement errors can be dynamic oscillation of te load of eac weel due to movement. Te transversal dynamic oscillations leads to different loads on eac weel in axle but not add te error to te common axle s weigt measurement results. Most sufficient seems te longitudinal oscillation because it leads to additional vertical force (or decreasing of it) to bot weels in axle simultaneously. For correction of tese errors we assume tat relative value of friction force, calculated by decomposition of te signal on even and odd parts [30], is dependent on footprint lengt and, respectively, it canges due to dynamic oscillations. From oter and, witout oscillations it is assumed to be in proportion wit te amplitude of relative derivative of te signal (rate of grow) (a) left rigt 0.06 friction left friction rigt (b) derivative left 0.04 derivative rigt Figure 9. (a) FOS signals from one axle, and (b) relative friction and derivative components for te same signals An example of tese components of te signals is presented on Figure 9 for bot left and rigt weels. As it is seen from te Figure 9, te friction and derivative of weigt component sligtly differs to eac oter by te amplitude. If te P max is te amplitude of weigt component s signal P(t) and DP max is te amplitude of weigt component s derivative dp(t)/dt, ten we can name te DP N value as normalized derivatives amplitude: DP max max DP N =, DF, Pmax SFmax = F N (6) SF max were = max( SF ( t )) min( SF ( t )), (7) F max is te maximum of friction component F(t) see Figure 9(b), and SF ( t ) = t 0 F ( τ )dτ is te integrated friction component s function. Te idea of errors correction is in te equalization of bot components for eac weel in axle, wen scaled friction component again add to weigt component to obtain modified signal S mod (t) witout influence of dynamic oscillations: S mod ( t ) = P( t ) + DP DF N N F( t ). (8) Ten modified signal (8) again must be under decomposition on weigt (even) and friction (odd) components for weigt calculation according to expression (4). Te results of application of proposed approac are reflected in Table 3. 75

80 Transport and Telecommunication Vol. 4, No, 03 Te accuracy of weiging te veicle in motion depends on many factors, but te main errors in te existing systems consist of te following: non-linearity and inertia of te sensor, termal effects, and inertial force of an oscillating veicle. Applying te algoritm of FOS signal processing wit te approximation of nonlinear caracteristics of te sensor (5) for a suitable range of temperatures, it is possible to calculate te following weigts of axes (Table 3): Table 3. Date: (Air Temperature + о С) Etalon axle s weigt (tons): Speed: 0 km/ No Parameter: st axle nd axle 3 rd axle 4 t axle 5 t axle Axle s weigt (tons) Error (%).% 5.77% -8.79% -.33% -6.84% Axle s weigt (tons) Error (%) 7.0% 7.03% -7.3%.4% 3.97% Speed: 0 km/ Axle s weigt (tons) Error (%).6% 5.0% -0.44% 0.74% 3.98% Axle s weigt (tons) Error (%) 4.87% 0.% -0.5% -6.0% -9.79% Speed: 50 km/ Axle s weigt (tons) Error (%) 8.9% -.0% -8.83% -6.80% -7.5% Axle s weigt (tons) Error (%) 6.% 5.5% -5.89% -8.8% -9.70% Speed: 70 km/ Axle s weigt (tons) Error (%) 7.95% -0.48% -6.34% -0.5% -.59% Axle s weigt (tons) Error (%) 4.37% -.75% -8.6% -.38% -3.0% Speed: 90 km/ Axle s weigt (tons) Error (%) -.9% 5.69% -5.79%.39% -0.9% Axle s weigt (tons) Error (%) 5.8% 5.87% -6.5% 0.76% -0.48% As it can be seen from te table (Table 3), most preferred for measurements are te velocity ranges from 70 km/ and above, wen te measurement errors of te load on te axle does not exceed 0%, wic is consistent wit te problem of pre-selection of overloaded veicles. Tis level of errors as te reason due to vertical oscillations of te dynamic motion of te veicle, wose amplitudes are smaller at iger speeds. By te properties of eac individual sensor, te calibration of FOS sould be conducted twice: firstly, in te laboratory (load caracteristics in te temperature range from -0 o C to +30 o C), and, secondly, after installing te sensor in te road surface wit veicles wit a standard load usage. 6. Conclusions Te model of seismic sensor excitation by automotive tyre footprint pressure is suggested. It is based on description of a footprint by certain discrete array considered as an equivalent source of a transient seismic signal. Proclaimed structure of te model is not contrary to widespread ideas of a motorcar weel interaction wit a road pavement. Te model is displayed te temporal form of tat pulse, wic is described by a convolution matrix reflected te interference of Raleig waves from different parts of te array. Te appropriate computing procedures ave been realized in Matlab. Resulted pulse forms may be used as a starting material for formulation of te target inverse problem in designing of WIM systems for estimation of loads on individual axles of a car. Te priori knowledge of te values and suspension stiffness of te tire, or one of tem, is need for successful application of te metod based on differential equations modelling. Unfortunately, te problem of determination of rigidity coefficient from te results of te optimisation as not been resolved, so we need to use its values from [4] or oter sources now. Because of a weigted least square algoritm 76

81 Transport and Telecommunication Vol. 4, No, 03 of global optimisation is used te processing times of WIM sensor signals are relatively long. It may be mentioned as some disadvantage of tis approac. But tis problem sould be solved by parallel processing of te signals from eac axle of te veicle. Dynamic WIM systems including FOS and reference database are not in using yet. Te obtained results ave to design tat system for determination of te overloaded veicles in traffic flows. Te use of te reference systems, wic are different in te structure, may increase te probability of te correct decision-making. One of te sounding problems is to create a set of te reference systems for different models of veicles. Te creation of suc a set for te large number of veicles is rater expensive venture. In tis connection, te development of te automated systems tat collects te statistical data and supports te reference identification systems of te overloaded veicles is really actual problem. To increase te accuracy of axes weigt in te motion measurement it is necessary to select te area of data representation, were te static and dynamic components of te weigt can be separated, let to be able to remove te dynamic components from te signal. Acknowledgements Tis researc was granted by ERDF funding, project Fibre Optic Sensor Applications for Automatic Measurement of te Weigt of Veicles in Motion: Researc and Development (00 0), No. 00/080/DP/...0/0/APIA/VIAA/094, References. Jacob B., Feypell-de la Beaumelle, V. (00). Improving truck safety: Potential of weig-in-motion tecnology. IATSS Researc, 34, Sengyao Jia et al. (00). Signal acquisition and processing of te moving veicle weigting system. WSEAS Trans. on Signal Proc., 3(6), Dorleus J. et al. (009). A fibre optic seismic sensor for unattended ground sensing applications. ITEA Journ., 30, Mimbela, L. E. Y., Klein, L. A. (005). Summary of veicle detection and surveillance tecnologies used in intelligent transportation systems. New Mexico: New Mexico State University. 5. Mazurek, B. et al. (00). Assessment of veicle weigt measurement metod using PVDF transducers. J. of Electrostatics, 5 5, US Dept. of Commerce. (997). Development of fibre optic dynamic WIM systems. Final Report. USA: Nat. Tecn. Inf. Service. 7. Seriff, R E., Geldart, L. P. (985). Exploration seismology, V.. Cambridge: Cambridge Univ. Press. 8. Tatom, F. B., Herndon, G. W. (Feb. 004). US Pat. No 6,69,567 B. USA. 9. Tarasik, V. P. (006). Teory of automobile motion. Saint-Petersburg: BHW. (In Russian) 0. Osinovskaya, V. A. (006). About evaluation and prediction of automobile road vibrations. (In Russian), Retrieved 006, from ttp://science-bsea.narod.ru/006/story_006/osinovskaya_vopros.tm. Mimbela, L.-E. Y., Pate, J., Copeland, S., Kent, P. M., Hamrick, J. (April 003). Applications of Fibre Optic Sensors in Weig-in-Motion (WIM) Systems for monitoring truck weigts on pavements and structures. Final report on researc project. New Mexico State University.. Levin, M. A., Fufayev, N. A. (989). Teory of deformable rolling weels. Moscow: Science. (In Russian) 3. Moazami, D., Muniandy, R., Hamid, H., Md.Yusoff, Z. (0). Effect of tire footprint area in pavement response studies. International Journal of te Pysical Sciences, 6(), Balabin, I. V. (Ed.) (985). Automotive and tractor weels: Handbook. Moscow: Masinostroenie. (In Russian) 5. Fernando, E. G., Musani, D., Dae-Wook, P., and Liu, W. (006). Evaluation of Effects of Tire Size and Inflation Pressure on Tire Contact Stresses and Pavement Response. Project Texas, USA: Transportation Institute, College Station. 6. Smit, N. D. (004). Understanding Parameters Influencing Tire Modelling. Colorado, USA: Colorado State University. Formula SAE Platform. 77

82 Transport and Telecommunication Vol. 4, No, Batenko, A., Grakovski, A., Kabaskin, I., Petersons, E., Sikerzicki, Y. (0). Weigt-in-Motion (WIM) Measurements by Fiber Optic Sensor: Problems and Solutions. Transport and Telecommunication, (4), Peters, B., Koniditsiotis, C. (000). Weig-In-Motion Tecnology, Intermediate Report, Austroads Inc., No. AP R68/00. Burwood Higway, Australia: ARRB Transport Researc Ltd. 9. Loo van, F. J. (00). Project WIM-Hand, st interim report, DWW-Publication: IB-R-0-09, Road and Hydraulic Engineering Institute, DG Rijkswaterstaat. 0. Busman, Rob, Pratt, Andrew J. (00). Weig In Motion Tecnology Economics and Performance. In Proc. of NATMEC 0. Carlotte. Nort Carolina: Warren Publising.. Giallorenzi, T. G., Bucaro, J. A., Dandridge, A., Siegel, G. II., Jr., Cole, J. II., Rasleigand, S. C., Priest, R. G. (98). Optical Fiber Sensor Tecnology. IEEE Journal of Quantum Electronics, QE-8, pp Sceuter, F. (997). General Guide for Weiging wit Portable Weel Load Scales: HAENNI, Document P 96. Maryland, Baltimore: Interscience Publisers. 3. Taylor, B., Klasinsky, R. (995). New Application for Weig-In-Motion Tecnology. Traffic Tecnology International. Surrey, England: UK & International Press. 4. Sivuin, D. (005). General Pysics Course. In Vol. I, Mecanics, 4 t Edition (560 p.). Moscow: FIZMALIT; MFTI Publ. (In Russian) 5. Widrow, B., Stearns, S. (989). Adaptive Signal Processing. Moscow: Radio and Svjaz. (In Russian) 6. Sakun, P., Sikerzicky, Y. (0). Adaptive metod s applications for te identification systems in dynamic weiging. In Proceedings of te t International Conference Reliability and Statistics in Transportation and Communication (RelStat ), 7 0 October 0 (pp ). Riga, Latvia: TTI. 7. Malla, Rames B., Sen, Amlan, Garrick, Norman W. (008). A Special Fiber Optic Sensor for Measuring Weel Loads of Veicles on Higways. Sensors, 8, Retrieved April 4, 008, from MDPI database on te World Wide Web: ttp:// 8. SENSORLINE GmbH. (00). SPT Sort Feeder Spliceless Fiber Optic Traffic Sensor: product description. Sensor line, GmbH. Retrieved January 7, 0, from ttp://sensorline.de/ome/ pages/downloads.pp 9. O Brien, E. J., Jacob, B. (998). European Specification on Veicle Weig-in-Motion of Road Veicles. In Proceedings of te nd European Conference on Weig-in-Motion of Road Veicles, 998 (pp. 7 83). Luxembourg: Office for Official Publications of te European Communities. 30. Mesco, A. (984) Digital Filtering: Applications in Geopysical Exploration for Oil, v.. Budapest: Academiai Kiado. 78

83 Transport and Telecommunication, 03, Volume 4, No *** Autors index Transport and Telecommunication, 03, Volume 4, No Transport and Telecommunication Institute, Lomonosov, Riga, LV-09, Latvia Autors' index Adamos, G. 0 Antoniou, C. Crocco, F. 45 Eboli, L. 45 Elmedany, W. M. 3 Fakr, M. W. 3 Grakovski, A. 57 Gyftodimos, E. Hussain, M. R. 3 Kabaskin, I. 57 Kasyanik, V. 39 Kondyli, A. Krasnitski, Y. 57 Lykogianni, G.-M. Mazzulla, G. 45 Natanail, E. 0 Pavlyuk, D. 9 Sut, V. 39 Sikerzitski, Y. 57 Truacov, V. 57 Zainal, A. 3 79

84 Transport and Telecommunication, 03, Volume 4, No *** Personalia Igor V. Kabaskin (born in Riga, August 6, 954) Vice-rector for Researc and Development Affairs of Transport and Telecommunication Institute, Professor, Director of Telematics and Logistics Institute Education & Scientific and university degrees: P.D. in Aviation (98, Moscow Institute of Civil Aviation Engineering), Dr.Sc.Habil. in Aviation (99, Riga Aviation University), Member of te International Telecommunication Academy, Member of IEEE, Corresponding Member of Latvian Academy of Sciences (998) Publications: More tan 390 scientific papers and 67 patents Fields of researc: Information tecnology applications, operations researc, electronics and telecommunication, analysis and modelling of complex systems, transport telematics and logistics Irina V. Yatskiv (born in Krasnoyarsk, Russia) Vice-rector of Transport and Telecommunication Institute, Professor of Computer Science Department Education: Riga Civil Aviation Engineering Institute, studied computer science and obtained an engineer diploma in 98 Scientific and university degrees: Candidate of Tecnical Science Degree (990), Riga Aviation University, Dr.Sc.Ing. (99) Scientific activities: Member of Classification Society of Nort America, Member of Latvian Simulation Modelling Society, Director of Latvian Operation Researc Society, te leader of project BSR INTERREG III B Programme InLoC, and projects connected wit transport system simulation in Latvia Publications: More tan 70 scientific papers and teacing books Fields of researc: Multivariate statistical analysis, modelling and simulation, decision support system Abdul Jalil M. Zainal Scientific and university degrees: Dr. Zainal olds a B.Sc. degree in Industrial Engineering from te University of Texas in USA, a P.D. in Remote Sensing from Imperial College/University of London (993), and a postgraduate diploma in Executive Management Leadersip from te University of Columbia (997). Dr. Zainal as part-time academic experience in teacing M.Sc. courses at AGU, in addition to supervising M.Cs. teses, and was te external examiner on oter occasions. Dr. Zainal is also te winner of te 990 Crown Prince Award for te best Scientific Researc Scientific activities: He is te CEO of STS (Spatial Tecnology Solutions) responsible for business planning, and managing te implementation of projects. His experience ranges from business strategy development and marketing, to products development and project management in te area of geo-spatial tecnology & telematics involving remote sensing, GPS and GIS. Prior to STS, Dr. Zainal was te CEO of Geomatec (commercial arm of Barain Centre for Studies and Researc, BCSR) until December 00. Dr. Zainal was responsible for rolling out Geomatec as a commercial arm of BCSR in 00, and in successfully transforming it from a small researc-oriented department into a self-financing business wit over 70 staff. From is position as a CEO, Dr. Zainal as establised a strong working relation wit many of te ITC/GIS users and stakeolders, and as establised links wit government, academic institutions, private and individuals alike Fields of researc: One of is major researc acievements as a project director as been te implementation of te Marine Geograpical Information project (MARGIS-II) Publications: Publication of te Marine Atlas Book of Barain in May 009. Dr. Zainal is te coeditor of te book. He as several publications in international scientific journals (mainly in te area of environmental applications), and participated as a speaker in many local and International conferences and seminars 80

85 Transport and Telecommunication, 03, Volume 4, No *** Personalia Marwan Raje Hussain Scientific and university degrees: Holds a B.Sc. in Computer Engineering from University of Barain, Barain, June 006. Currently e is doing is M.Sc. in Management Engineering in te same university. He works as a consultant in Telematics Department at STS (Spatial Tecnology Solutions) in Barain. Prior to STS, Marwan was te Acting Manager of AVL & Telemetry Solutions of Geomatec (commercial arm of Barain Centre for Studies and Researc, BCSR) since August 006 until December 00 Scientific activities: One of is best projects in B.Sc. was about Te Design of a PC Based Electrocardiogram (ECG) Device and Traffic Controller Design using VHDL (Hardware Description Language) Fields of researc: Telematics, AVL & Telemetry Solutions, wireless communication engineering, and FPGA based system design Publications: Some publications in ITS Wael Moamed Elmedany Scientific and university degrees: He olds a P.D. degree in Electrical Engineering from Mancester University, UK, July 999 Scientific activities: He is te founding and managing editor of International Journal of Computing and Digital Systems (IJCDS). He as been Awarded te prize of excellence in Designing and Developing e-courses at te Universities of Gulf Cooperation Council, 00, and e nominated and is biograpy was included in te 9 t and 7 t Editions of Wo s Wo in te World, 0 and 00. Elmedany is also IEEE member, as well as a member of IASTED Tecnical Committee on Telecommunications, Canada, He is also a reviewer in many international journals and conferences, and acts as cairperson in many conferences Fields of researc: His researc interest in te applications of ASIC design, FPGA, embedded systems, and reconfigurable computing Publications: Elmedany as over tirty publications, and e attended several national and international conferences, and worksops Moamed Waleed Fakr Scientific and university degrees: P.D. from University of Waterloo, Canada, Electrical Engineering Department, 993. Joined NORTEL from as a researcer in te speec recognition researc lab in Montreal. In 999 e Joined te Arab academy for science and tecnology were e is a professor in te computer engineering department, currently on leave at te University of Barain Dr. Fakr researc interests are in signal processing, pattern recognition and multimedia security were e as publised more tan 50 papers Scientific activities: Dr. Fakr is a reviewer for Springer Multimedia Tools and Applications journal and Elsevier Systems and Software Journal. Dr. Fakr is also a member of te editorial board of te international journal of Artificial intelligence & applications (IJAIA) and te international journal of Multimedia & its applications (IJMA). Dr. Fakr is a co-founder of te Arabic Language Tecnology Center (ALTEC: ttp:// and a member of IEEE since 98 Publications: More tan 50 papers Constantinos Antoniou (born in Atens, Greece) Assistant Professor in te National Tecnical University of Atens, Greece Education: Diploma in Civil Engineering from NTUA (995), a M.Sc. in Transportation (997) and a P.D. in Transportation Systems (004), bot from MIT Fields of researc: His researc focuses on modelling and simulation of transportation systems, Intelligent Transport Systems (ITS), calibration and optimization applications, and road safety Publications: He as autored more tan 35 scientific publications, including 37 papers in international, peer-reviewed journals, 95 in international conference proceedings, a book and 5 book capters 8

86 Transport and Telecommunication, 03, Volume 4, No *** Personalia Alexandra Kondyli (born in Atens, Greece) Present position: Postdoctoral associate at te Transportation Researc Center, University of Florida Education: P.D. in Transportation Engineering (009, University of Florida), M.Sc. in Transportation Engineering (005, University of Florida), Diploma in Rural and Surveying Engineering (003, National Tecnical University of Atens) Publications: 7 scientific papers Researc fields: Traffic operations, simulation, traffic flow teory, driver beavior Elias Gyftodimos Researc collaborator at National Tecnical University of Atens, Scool of Rural and Surveying Engineering, Laboratory of Transportation Engineering P.D. in Computer Science (006, University of Bristol, UK). Degree in Informatics (000, National & Kapodistrian University of Atens, Greece) Researc interests: Macine Learning, Data Analysis Publications: 3 scientific papers Georgia-Maria Lykogianni (born in Ioannina Greece, December, 989) 007 0, Diploma of Survey and Rural Engineering, National Tecnical University of Atens (NTUA), Greece Master student of Transport and Geoinformation Tecnology, Kungliga Tekniska Högskolan (KTH), Stockolm, Sweden (0 today) Diploma tesis: Investigation of te performance of Dynamic estimation of Origin-Destination matrices Giannis Adamos (born in Volos, Greece, May 4, 978) Researc Associate, Hellenic Institute of Transport, Centre for Researc & Tecnology Hellas, Tessaloniki, Greece; Researc Associate, Transportation Engineering Laboratory, University of Tessaly, Volos, Greece Education: 5-year Diploma in Civil Engineering ( ), M.Sc. in Applied Engineering & Systems Simulation ( ), P.D. Candidate (008 today) Fields of researc: Road safety, simulation of driving beaviour, Intelligent Transport Systems, public transport, design of urban transport intercanges, intermodal passenger and freigt transportation Publications: More tan 30 in scientific journals, conference proceedings and books 8

87 Transport and Telecommunication, 03, Volume 4, No *** Personalia Eftiia Natanail, P.D., M.Sc. (born in Tessaloniki, Greece) Position: Head of te Transportation Engineering Laboratory, Ass. Professor in Transportation Systems Design and Evaluation, Department of Civil Engineering, University of Tessaly, scientific consultant at te Hellenic Institute for Mobility and Transportation of te Centre for Researc and Tecnological Development HELLAS Education: Diploma in Surveying Engineering, Aristotle University of Tessaloniki (988), M.Cs. in Civil Engineering (Transportation), University of Miami, USA (99), P.D. in Civil Engineering Aristotle University of Tessaloniki (996) Scientific activities: Committee Member of RESTRAIL, EU, ECOMOBILITY, ECOCITY, EU, Hazardous Material Transportation (AT040), TRB, national representative of COST-TU004, Support Framework Business program and Foresigt Tecnology, Greece, European Tematic Network ROSEBUD Publications: More tan 80 scientific papers publised in journals and conference proceedings and book capters in 5 books Fields of researc: Transportation planning and design, intelligent transportation systems, multicriteria evaluation and optimization Laura Eboli Scientific and university degrees: P.D. in Environmental planning and tecnologies in 008 from te University of Calabria, Italy. Se olds a master s degree in transport systems management from te University of Calabria. Se graduated in Environmental Engineering in 00 from University of Calabria Assistant Professor of Management of te Transportation Systems. Since 004 se as cooperated to te didactic and researc activities in te Department of Land Use Planning of te University of Calabria. Se teaces transportation management, and traffic teory at te Faculty of Engineering of te University of Calabria Fields of researc: Transportation demand and supply analysis, transit system wit particular interest in te measure of service quality, analysis of interaction among land use, environment and transports, and road safety. Te major part of te researc as focused on transit service quality, wic was also te subject of er P.D. tesis Publications: Autor of about 60 papers, edited in Italian and international literature. Her publications ave appeared in transportation journals, including te Transportation Planning and Tecnology, Journal of Public Transportation, European Transport, Transport Reviews, Transport Policy, International Journal of Sustainable Transportation Gabriella Mazzulla Scientific and university degrees: P.D. in Road Infrastructure and Transportation System, acieved in 00 from University Federico II of Naples, Italy. Se graduated in Civil Engineering, wit orientation in Transportation Engineering, in 995 from Calabria University, Italy Assistant Professor of Fundamental Elements of Transportation and Transit Systems. Since 996 se as cooperated to te didactic and researc activities in te Department of Land Use Planning of Calabria University Fields of researc: transportation planning, specifically analysis, modelling, and estimation of travel demand, design of transportation networks and simulation of traffic flow, and environmental impact appraisal of transportation facilities, wit particular regard to atmosperic and noise pollution from road traffic. Se contributed to some national and international researcer projects as coordinator or member of te working group Publications: Autor of about 00 publications, edited in Italian and international literature. Her publications ave appeared in transportation journals, including te Transportation Planning and Tecnology, Journal of Public Transportation, European Transport, European Journal of Operational Researc, Transport Reviews, Transport Policy, International Journal of Sustainable Transportation 83

88 Transport and Telecommunication, 03, Volume 4, No *** Personalia Federica Crocco Scientific and university degrees: P.D. in Environmental planning and tecnologies in 00 from te University of Calabria, Italy. Se undertakes researc in freigt transport and logistics and in road safety. Fields of researc: Her researc interests are also in te general area of intelligent transportation systems wit focus on advanced traffic signal control systems, integrated control of transit and traffic operations, traveler information systems and information tecnology applications to transportation. Publications: Dr. Crocco is autor of about 0 papers, publised in proceedings of national and international conference, in national journals and in volumes about territorial and transportation planning Vasili Sut (born in Magnitogorsk, Russia) Education: Vasili Sut graduated from te Taganrog Tecnical Institute in 97. From 97 till 996 e ad been working at te Brest production association as a cief of special equipment design office for computer systems. Since 996 Vasili Sut works at te Brest State Tecnical University Scientific and university degrees: Head of transport laboratory in Brest State Tecnical University, Assoc. Professor in Department of Intellectual Information Systems, USSR Honorary Iinventor Fields of researc: Area of scientific interests: intellectual and adaptive traffic systems, grap teory, optimizing problems of discrete matematics Publications: 50 publications and 5 patents Valery Kasyanik Education: M.Sc., P.D.-student, assistant of teacer in department of intellectual information tecnologies and robotics laboratory. Area of scientific interests: mobile robotics, artificial intelligence Publications: Autor of 6 articles Alexander V. Grakovski (born in Riga, November 5, 96) Scientific and university degrees: Professor, Dr.Sc.Ing., Director of Bacelor s Academic Study Programme in Telecommunication Systems and Computer Networks, Head of Electronics and Telecommunications Department, Dean of Computer Sciences and Telecommunications Faculty of Transport and Telecommunication Institute University study: Riga Institute of Civil Aviation Engineers ( ) P.D. in Electrical Engineering (989, Moscow Energetic Institute), Dr.Sc.Ing. in Electrical Engineering (99, Riga Aviation University) Fields of researc: Computer modelling, caos and non-linear dynamics, fractals, transport veicle route optimization, digital image and signal processing in telecommunications, navigation, transport telematics Publications: 56 scientific papers and 7 teacing books 84

89 Transport and Telecommunication, 03, Volume 4, No *** Personalia Yury Sikerzitsky (born in Ukraine, ) Professor of Transport and Telecommunication Institute University study and Scientific and university degrees: Riga Civil Aviation Engineering Institute ( ); Candidate of Science in Engineering (98), Riga Tecnical University; Associated Professor s Certificate in Engineering (988), Riga Institute for Civil Aviation Engineers; Doctoral Degree in Engineering (99), Riga Aviation University; Doctoral Habilitation Degree in Engineering (993), Riga Aviation University; Professor s certificate (995), Riga Aviation University; Professor s diploma (007), Riga Tecnical University, Transport and Telecommunication Institute Publications: More tan 90 scientific publications, textbooks Scientific activities: Adaptive signal processing, Radio receiver devices, Video tecniques, Teory of telecommunication systems Yury A. Krasnitsky (born in Penza, Russia, Apr., 938) Professor, Dr. Habil.Sci.Ing., Professor of te Telecommunication Department Education: Leningrad Precise Mecanics and Optics Institute (955 96) Scientific degrees: Candidate of Tecnical Sci. (969, Riga Institute of Civil Aviation Engineering); Dr. Pys.-Mat. Sci. (989, Leningrad Main Geopysical Observatory), Dr. Habil.Sci.Ing. (99, Riga, Aviation University) Fields of researc: Electrodynamics and antennas, wave propagation, natural electromagnetic penomena, signal computer modelling and processing Publications: Near 50 scientific papers, 5 patents, 0 tutorials and teacing books Victor Truacov (born in Alma-Ata, Kazastan, Febr. 8, 944) Dr.Sc.Ing., Senior Lecturer (Docent) of te Telecommunication Department Education: Degree of Radio Engineering from Riga Institute for Civil Aviation Engineers (96 967) Scientific degrees: Candidate of Tecnical Science (973, Riga, Institute for Civil Aviation Engineers); Doctor of Science Degree in Engineering (99, Riga, Aviation University) Fields of researc: Electrical Circuit Teory, Matematical Simulation and Optimization of te Systems, Signal and Image Processing Publications: Near 40 scientific papers, 0 tutorials and teacing books Dmitry V. Pavlyuk (born in Saratov, Russia, July 6, 980) Docent of Matematical Metods and Modelling Department Education: Saratov State University, Department of Mecanics and Matematics, Information Systems (in Economics) (00) Scientific and university degrees: Candidate of Economic Science Degree (005) on Matematical Metods in Economics Publications: 5 scientific papers and abstracts for conferences Fields of researc: Econometric analysis (stocastic frontier models; spatial models), statistics 85