Transactions on the Built Environment vol 18, 1996 WIT Press, ISSN

Size: px
Start display at page:

Download "Transactions on the Built Environment vol 18, 1996 WIT Press, ISSN"

Transcription

1 A Decision Support System for the operative planning of railway service on a national scale F. Russo% G. Beltrami "Faculty ofengineering, Reggio Calabria University, Via Emilio Cuzzocrea, Reggio Calabria, Italy ^Transport Area, Italian Railways (Ferrovie dello Stat S.p.A.), Milan, Italy Abstract In order to analyse market requirements, to improve traffic returns and to reduce operating costs of the long distance services, the Italian Railways (FS) are defining and testing a decision support system named SASM (System of Analysis and Simulation for Operative Marketing) that supports the activities of market quantification and segmentation, competitive framework definition and effects forecasting. The main emphasis of this paper is in presenting the general architecture of the SASM and its innovative modelling contents. The need to take into account the scheduling of rail and alternative services has required the extension of traditional demand, supply and assignment models to include an explicit within-day dynamic dimension. The first part of the paper describes the objectives of the DSS, the main functions with the overall functional architecture and the main components. In the second part of the paper, the used spatio-temporal network, the system of demand models and the assignment model are analysed. In the third part the main functions of the interface are described. Finally, a preliminary application of the system on a test rail corridor is described.

2 220 Computers in Railways 1 General characteristics of the system 1.1 Objectives and functions The SASM is currently being set up at the Transport Area of the Italian State Railways to aid decision-making for corridor operative marketing plans relative to long-distance railway services: intercity, interregional/regional and night (SINTRA^; Cascetta, Nuzzolo & Beltrami*; Nuzzolo, Russo & Crisalli*). The considered decision variables concern the operating characteristics (paths, stops, timetables and train compositions) and sale prices of the services. As what regards the Marketing Plan in general terms, the support system is particularly useful in the current situation analysis phase and in the definition phase of marketing strategies through the use of the functions of: supply analysis and comparison of characteristics for railway services and competitive modes (car, bus, plane); effect forecasting, in terms of traffic and returns for the railway system, of variations in service characteristics and service prices. 1.2 General functional scheme and main components Performances of railway and competitive mode/services, passenger "loads" and related returns of each single train for a design scenario are computed through a sequence of phases consisting of (see Figure 1): formulation of the design scenario, particularly by the definition of operating characteristics and prices of railway services; implementation of supply models of the design scenario; calculation of probabilities of each alternative rail/service/run/class configuration for each demand segment and for each O/D pair, by means of a system of choice models; computation of the passenger number d^(s,ser,cl) of users of demand segment s on O/D pair, that use rail service ser and class cl. computation of new O/D flows on each scheduled run of the railway services using the assignment model; determination of railway revenues. The main SASM simulation components are thus as follows: 1) Supply models, that allow to calculate the level of services attributes of the various services/modes for O/D pairs and to carry out the assignment procedures : a) Infrastructural road network model and terminals, consisting of : inter-provincial, inter-regional and national roads graph, with zone centroids and connectors to the network nodes, terminal nodes of railway, plane, bus and ferry, and the relative access arcs; link characteristics, in particular travel time and cost; b) Railway infrastructural network model;

3 Computers in Railways 221 c) Diachronic network models of railway services (obtained from train timetables for intercity, regional/inter-regional and night services), of air services, of inter-regional bus services and of corridor interprovincial bus services (see section 2.2); d) Network model of access/egress from centroids to terminals 2) Demand models (see section 2.1); 3) Assignment models (see section 2.3). Others main components are: 4) User interface functions: initialisation, making of scenarios, comparison of alternative modes and services, calculation of effects (see section 3); 5) System input data: the chief input data consist of traffic zone data, supply data (graph and link characteristics for infrastructural networks) and demand data. DESIGN SCENARIO r Mode I.O.S ^/service characteristics Figure 1 - Functional architecture of the simulation 2 The system of models 2.1 The demand model For a correct simulation of user behaviour in response to the variations in service characteristics and thus for a better evaluation of returns and of flows on the network, it is necessary to represent explicitly the services with run departure times and to use demand models which simulate user choices

4 222 Computers in Railways in each market segment. Market segmentation takes into account the purpose, the origin or destination target time, the car availability, the household income and the number of users travelling together. Several modal choice models have been proposed either to simulate the possible modal split changes resulting from short-medium term actions such as variation of the fare structure; for a detailed analysis of the literature, see Miller & Fan^ and Cascetta, Nuzzolo & Biggiero\ In any case it can be concluded that such models, albeit complex, are not suitable for simulating the impacts on demand of the above-described operative marketing choices. The proposed system of demand models (Cascetta, Nuzzolo & Biggiero^) recognises explicitly that the mode "train" is in fact a "bundle" of different services varying by their performances and price. It also recognises that the "attractivity" of a scheduled service depends on their convenience with respect to desired arrival/departure time and the accessibility of their terminals. In other words the demand model allows, for each demand segment, to reproduce mode and service choices on every O/D pair and, in the case of scheduled services (train, plane, bus), the choice of run and boarding/alighting stations, so it allows to compute individual train loads. The model assumes that the choice alternatives faced by user when undertaking a trip have various dimensions (Figure 2): primary mode i.e. that used for the longest part of the journey and, in the railway case, service and class of the primary mode; run with related access and egress terminals; access and egress mode from the terminals. Such complex choice structure will be modelled by means of a Tree-Logit functional structure (Ben Akiva & Lerman*; Daganzo & Kusnic^). In the proposed model; the probability of choosing a single alternative in the most complex case is given by equation p(pm,ser,r,tr^,tig,cl,m^,mg /O,D,s) = = p(pm/ O,D,s) { p(ser / O,D,s,train) p(r,tr^,tr^ / O,D,s,(ser),pm) p(cl/o,d,s,r,trg,trg,ser,pm)"p(m^,nig /O,D,s,r,ti;,tr^,(cl,ser),pm) } where: pm primary mode ser rail service (intercity, regional/interregional, night) r run tfa, tfe access and egress terminal cl service class (I, II) nia, nie access and egress mode O, D origin and destination s market segment

5 Computers in Railways 223 O, D,s in PRIMARY MODE Night SERVICE T--T T-- T--TT T--TT T ACCESS AND EGRESS MODE Figure 2 - The tree-logit model representation 2.2 The diaehronic network model This section reports the approach used for the description of the service schedule with a diachronic network model and for generating "feasible" terminal-run alternatives, given a mode/service, an O/D pair, a market segment with an arrival or departure target time. The network model used (Figure 3) represents each run of each line with the explicitation of the variable time; this model is called "diachronic network" (Nuzzolo & Russo*'^). The general model of the diachronic graph consists of two subgraphs: one relates to run representation and the other, which is composed of various subgraphs, at least one per zone centroid, in relation to the structure of the matrix O-D, segmented by purpose and destination (arrival) target time (DTT) or origin (departure) target time (OTT). In specific cases, as this DSS, it is possible to use only the part referred to the runs. The proposed diachronic network model allows to adapt usual network algorithms to the generation of sequences of runs connecting an origin to a destination in function of OTT or DTT and for the computation of their attributes (on board travel time, access/egress time, schedule delays).

6 224 Computers inrailways Transactions on the Built Environment vol 18, 1996 WIT Press, ISSN A stop axis k access \ \ origin % centroid Figure 3 - Example of diachronic network 2.3 The assignment model For the assignment of the O/D matrix to the railway service, SASM uses a Stochastic Network Loading assignment model, with explicit path enumeration. For a given O/D pair and a given market segment, the path choice set includes three service alternatives (intercity, interregional/regional and night), two runs for each service (the last scheduled run before the target time and the first scheduled run after the target time) and two class alternatives (I and II class) for each run. The path choice model is a part of the system of choice models described in section 2.1. The diachronic network model and the system of demand models described in the previous section allow to carry out, the path choice set and the relative attributes for the path choice models. The path generation and loading procedure for a given railway service, origin and destination pair, a given demand segment and target departure or arrival time, is a follows: - identification of "feasible" access and egress terminals according to preassigned centroid-terminal incidence matrices; - identification of the set of alternative diachronic paths; - linking of diachronic and access (egress) paths;

7 calculation of the probability of using each alternative; train loading of each alternative. Computers in Railways 225 Transactions on the Built Environment vol 18, 1996 WIT Press, ISSN User interface The SASM has been realized in environment integrated for "Windows"; the system has been implemented as an application for Windows 3.1 and following, and it satisfies a detailed list of items related to the management of the memory and the speed of execution, depending on the hardware. In fact, the definition of the project work considers the complexity of the system of data and of the number of algorithms to manage, factors that bias the performances of the system in conclusive manner. The DSS consents the selection of the following macrofunctions (Figure 4): FILE UTILITY M; \IN Dj* ita 1 SELECT REFEIIENCE SCE> IARIO 1 RESTORE AND BAG KUP SCE> 7ARIO 1 SCEN^IRIOS I T DAE TAB LES 1 FAFIBS DIACHRONIC 1 NETWORKS 1 1 CALCULATION GRAPHIC OF TOOLS EFFECTS : i i i SELECT 1 ZONING PLAN SCENARIO 1 ANALYSIS AND COMPARISON IS INGLE SERVICE 1 ANALYSIS 1 MODES/SERVICE 1 ANALYSIS MARKET 1 ANALYSIS 1 EVALUATION: -MARKET SHARE -FLOWS -REVENUES 1 1 COMPARISON 1 SCENARIOS RAILWAY NETWORK ROAD NETWORK Figure 4 - Macrofunctions of the DSS. File utility Scenarios Analysis and Comparison Calculation of effects Graphic tools The unity of reference for the evolution of the functions is the SCENARIO. To distinguish the diverging configurations that could engage the system of transport according to contingent modifications on the base data, we need to insert the set of relative data into each individual configuration in a scenario of reference. Then with scenario the analyst qualifies the set of data that characterize completely the supply generated from the systems of transport considered. Through the macrofunction File Utility it is possible to:

8 226 Computers in Railways insert the main data relative to the demand and to the supply (zoning, infrastructural networks, timetables of the non railway services, O/D matrices for service and class, etc.) necessary for the simulations and for the following analyses; select a scenario of reference on which all the analyses will perform; - effect a series of operations of utility (Copy Files, Change directory, Backup or Restore of a scenario). Through the macrofunction Scenarios it is possible to: insert and/ or alter manually the schedules for each individual railway service (sub menu Time tables); build the matrix of the prices of the railway services (sub menu Fares) that could be defined for a kilo me trie band, for O/D pair, for day of the week, for hourly band and for an individual train; generate the diachronic networks. Through the macrofunction Analysis and Comparison it is possible, having selected a scenario of reference with the function File, to carry out: an analysis of the supply relative to a specific railway service (Single Service Analysis); a compared analysis (Mode/Service Analysis) among different modal alternatives (car, aeroplane, bus) and alternative railway services (intercity, interregional and night) in terms of characteristics of the alternatives (times, costs, distances etc.). an analysis of the matrices of demand for railway service (Market Analysis). Through the macrofunction Calculation of Effects it is possible, having selected a scenario of reference with the function File and a scenario of plan, to carry out: an estimate of the O/D matrices of plan for railway service (Market Share); a simulation of the interaction supply-demand (loads of the trains) by means of a model of assignment to the different networks with visualization of the loads for train; an evaluation of the revenues for train; a comparison among Scenarios of Reference and of Plan in terms of market share, loads of the trains and revenues. Through the Graphic tools it is possible, having selected a scenario of reference, to enter a GIS environment to display the zoning, the railway and road infrastructural networks with relative characteristics. 4 Some elements of a test application The system of analysis and simulation described in the previous sections has being implemented on a personal computer with Pentium CPU/133MHz/16 Mb RAM and tested on the Turin-Milan-Venice railway corridor; this is one of the two fundamental lines in Italy, crossing four regions with approximately 100 long-distance trains per day. In this application, Italy was

9 Computers in Railways 227 divided into 359 traffic zones, 178 of which were in the area more directly affected by the sample corridor and 181 in the rest of the country. The road infrastructural network model covers the whole of Italy and has 1850 nodes and 6631 links. In particular, 441 railway terminals and 28 airports are represented. The railway infrastructural network model includes the railway lines of the State Railways nationwide; overall, the graph consists of 448 nodes and 1086 links. Diachronic service network models are of a considerable dimension: for example the regional/inter-regional services network consists of 11,000 nodes and of 41,000 links. Calibration of the path choice models is being carried out, using the data of an on-board train survey with about 10,000 interviews on the corridor. For example, for the business travel the first results of the calibration (Cascetta, Nuzzolo & Biggiero^) give the following utility function: where: Tb = on-board travel time (h) Cmh = on-board travel cost for medium-high income users (looolit) Cl = on-board travel cost for low income users (looolit) EA = early arrival time of the run with respect to the target time (h) LA = late arrival time of the run with respect to the target time (h) Ht = specific preference for intercity trains IcIL = specific preference for the 1st fare class in interregional/regional trains IcIH = specific preference for the 1st fare class in intercity trains The night service is not enabled on the corridor, so this service has not been considered. Table 1 - Path choice model for "business" (Cascetta, Nuzzolo & Biggiero^) coeff (3 t-st Tb %& a EA LA Ht c/L IcIH Orun c/af rho squared = 0.30 rho bar squared = 0.27 Using the calibrated demand models and the diachronic network of services, it has been carried out the assignment, obtaining the load of the trains. The passengers on board have been counted on some trains on the subsequent links, therefore it is possible to compare counted and model estimated flows. This comparisons gave satisfactory results, with differences minus than 10%. In figure 5 is reported the comparison for a train intercity Milano-Torino (IC642), referring to the T and 2 class passengers.

10 228 Computers in Railways Figure 5 - Comparison counted estimated flows (train IC642) References 1. Ben Akiva M. & Lerman S.R. Discrete choice analysis. MIT Press, Cambridge Mass, Cascetta E., Nuzzolo A. & Beltrami G. A system of demand and supply models for the marketing of intercity passenger services, Seventh World Conference on Transportation Research, Sydney, Cascetta E., Nuzzolo A. & Biggiero L. A within-day dynamic tree logit model for mode-service choice with explicit representation of time-table. Internal Paper Department of Transportation Engineering, "Federico II" University, Naples, Daganzo C. F. & Kusnic M. Another look at the Nested Logit model. Research report of the Institute of Transportation Studies, University of California, Miller E.J. & Fan K.S. Travel demand behaviour; survey of intercity mode-split models in Canada. Royal Commission on national passenger transportation, Minister of Supply and Services, Canada, Nuzzolo A. & Russo F. Un modello di rete diacronica per 1'assegnazione dinamica al trasporto collettivo extraurbano. Ricerca Operativa n. 67, pp , Nuzzolo A. & Russo F. An Equilibrium Assignment Model for Intercity Transit Networks. Triennal Symposium on Transportation Analysis, 1994a 8. Nuzzolo A. & Russo F. Departure time and path choice models for intercity transit assignment. Preprint IATBR-94, Cile. 1994b 9. Nuzzolo A., Russo F & Crisalli U. A decision support system for railway service planning. Euro Working Group on Transportation, pp BarceUona lo.sintra SASM - Sistema di analisi e simulazione per il marketing operativo Progetto esecutivo, Ferrovie dello Stato, 1994