Developing Measures of Public Transport Schedule Coordination Quality

Transcription

1 Developing Measures of Public Transport Schedule Coordination Quality Graha Currie 1, Leigh Broley 2 1 Professor and Chair of Public Transport, Institute of Transport Studies, Monash University, VIC, Australia 2 Research Student, Departent of Civil Engineering, Monash University, VIC, Australia 1 Introduction Despite the rhetoric and "big bucks", the governent is yet to provide an exeplary interchange "opportunity" anywhere in Brisbane. For exaple, if you wish to travel fro the city to St Lucia, there is no evidence of coordination of bus and rail or ferry in ters of tietables or routes Yeates (2002) Coordinating the tietables of fixed route public transport has long been an ai of public transport planners and the counity. A well coordinated tietable enables easy transfer between public transport services. This involves a balance between allowing enough tie to transfer between alighting to boarding services whilst also iniizing waiting tie during the transfer. Effective transfers can substantially increase the catchents of bus and rail service and enable network wide travel on public transport. Wong and Leung (2004) deonstrated that adjustents to schedules for the optiization of transfer ties on the Hong Kong Mass Transit railway could reduce passenger wait ties by between 43% and 73% for little iediate cost. Tietable coordination is one of any planning easures included in network integration strategies for public transport systes. Despite regular reference to policies which seek to iprove coordination there is little deonstrated evidence of iproveents resulting fro such policies. A ajor proble is that it is difficult to easure coordination and hence it is difficult to deonstrate the benefits of iproved coordination. Lack of bus rail coordination reains a coon coplaint of users of public transport throughout the world. This paper explores the issues surrounding the proble of schedule coordination on public transport networks. It takes the view that a ajor proble in planning for schedule coordination is the lack of a robust and objective easure of schedule coordination quality. Without a practical eans of quantifying coordination quality, there is no systeatic and defendable basis for deonstrating the benefits of iproved coordination or a eans of planning for optial approaches to iprove schedule coordination. This paper starts with a review of issues associated with schedule coordination to provide context. It then reviews approaches to easureent. A relatively new approach to easuring schedule coordination tered the Synchronization Quality Index is then outlined and the results of an application of this ethod described. A new developent of the ethodology tered the Synchronization Quality Ratio is then presented and results fro a sall scale application of this approach outlined. Soe conclusions on the next steps required to realize the potential benefits of the new approach are presented. 2 Issues in Schedule Coordination This section explores issues affecting the planning of public transport schedule coordination to provide context for the paper. 28 th Australasian Transport Research Foru Page 1

2 2.1 Mixed and Conflicting Objectives Adjusting schedules to optiise transfer ties is only one of any factors which authorities ai to achieve with schedule developent. Ceder et al (2001) point out that transport operators have nuerous objectives when setting tietables e.g. Miniising operator costs Miniising user wait ties Miniising vehicle and crew allocations Deterining appropriate frequencies based on deand Matching the schedule to crew working conditions. The iniisation of passenger wait ties can often conflict with a requireent to iniise vehicle and crew requireents since the forer requires that vehicles spend tie waiting whilst the latter requires faster speeds and shorter waiting. In any cases poor vehicle and crew utilisation can be ore a direct concern of operators than passenger waiting tie because it is a ore direct and tangible factor which affects financial perforance. 2.2 A Coplex Tie Space Issue While it can often appear desirable at a localised level to shift a particular scheduled arrival tie of bus A to better eet with train B, it can often be undesirable and ipractical on a network wide basis to do this. Shifting ties of a particular arrival eans that all other tiings on that particular bus vehicle trip are also adjusted. This can cause two probles: a. Other coordinated arrival ties along the route can be affected (both positively and negatively) b. The even headway interval between bus trips can be ade uneven. This can cause an uneven service perforance for those waiting at stops along the route and in extree cases cause uneven loadings and bunching of services. Schedule coordination is thus a coplex nuerical proble. It lends itself to coputational and atheatical research ethods. Indeed ost large public transport agencies eploy sophisticated scheduling software such as the HASTUS syste (Rousseau and Haer, 1993) to identify optial solutions which balance their needs for inial vehicle and crew inputs to the passenger requireents of schedules. Coplexity also acts to ake schedule coordination a very abstract and difficult to describe issue. User groups in particular can see tangible localised probles with schedules at a given station but do not perceive the network wide constraints on scheduling. Coplexity also akes transit planners focus schedule planning on a few high profile locations. Rarely are the wider ipacts of adjusting schedules to achieve better coordination at a particular location traded off against ipacts on a larger nuber of ediu and lower order interchanges. 2.3 The Constraints of Multi-Operator Planning Coputerised schedule planning systes such as HASTUS can trade off network coordination of schedules against vehicle and crew resource iniisation goals. However this is alost never a practical solution to schedule coordination for ost cities: Certainly in Australasia, no cities have bus and rail (and tra) schedules being planned by a single central ulti-odal scheduling syste Rather each city has a separate scheduling syste for rail, for bus (and for tra) More often there are any alternative scheduling systes being used by different bus copanies ranging fro anual to high quality coputer systes. In Melbourne there are 37 bus operators and at least 10 alternative ethods to scheduling. 28 th Australasian Transport Research Foru Page 2

3 The general approach to this issue has been to andate that all operators ust plan schedules with coordination in ind. These requireents are usually included in operating contracts. In Victoria for exaple, the Departent of Infrastructure has established guidelines for odal coordination whereby bus operators ust ake their best endeavours to accoplish tietable coordination. Little detail on what this eans is specified or detailed onitoring of outcoes apparent. Operators ay for exaple be faced with a very real trade off between additional costs in vehicle and crew requireents as a result of schedule changes which iprove network coordination. No foral approach to this issue appears to have been developed. In the absence of clear direction fro regulators and also because of the coplexity of the issue, schedule coordination is soewhat glossed over. Planning is undertaken but outcoes are not clear. 2.4 Schedule Coordination in Practice The research literature ters schedule coordination schedule or network synchronization. The difficulty of the task and the approach ost coonly taken is typified by the following quote fro the literature: Synchronization is the ost difficult task of transit schedulers and is currently addressed intuitively Ceder et al (2001) Public transport schedulers typically identify critical coordination nodes on the network and plan schedules around achieving reasonable coordination at these points. In any cases tietables involve tied transfers i.e. with connections to specific pairs of rail and bus services identified within the schedule. This approach is a practical and effective eans of targeting the coordination proble at ajor interchanges; however it tends to ignore the issue of coordination at second and third tier interchanges. The intuitive approach avoids the coplex task of balancing coordination on a network wide basis in favour of a pragatic attention to a few key points. Regulators of public transport are therefore faced with a difficult and coplex proble. Because ultiple approaches to scheduling are used no systeatic syste wide approach to optiising schedule coordination is feasible. This is exacerbated by the lack of an easy approach to easuring schedule coordination quality. 3 Measuring Schedule Coordination There are 2 ain approaches to easuring schedule coordination quality; passenger wait tie and the Synchronisation Quality Index (SQI). 3.1 Passenger Wait Tie The ost coon approach to easuring the quality of schedule coordination is to su passenger waiting tie. Voss (1992), Daduna and Voss (1995) and Ceder et all (2001) used an approach involving: Measuring passenger wait ties (including transfer walk tie) for transfers Weighting transfer ties by passenger volues Developing atheatical optiisation techniques which ai to iniise passenger wait ties and balance this against other schedule developent objectives. There are a nuber of probles with this approach: 28 th Australasian Transport Research Foru Page 3

4 Data Availability - Inforation on passenger travel volue is not always readily available particularly on a network wide basis and certainly not for individual transfers. While inforation on schedule tiings is generally readily available, passenger deand data on individual transfers requires considerable data collection resources. Miniising Wait Tie Can Be The Wrong Objective Although iniising transfer tie is an attractive objective, if ties are too short there is a danger that soe passengers, notably those with obility challenges, ay not be able to ake a given transfer. In practice it ay be appropriate to differentiate between: a. a iniu transfer tie, where soe passengers can ake the transfers b. a axiu transfer tie, where all passengers can ake the transfer; and c. an ideal transfer tie, which is a balance between a and b. In this case providing good coordination is an optiisation process which seeks a balance between ties a. b. and c. Miniising transfer wait tie is a siplistic approach when the tiing of transfers and the iplications for different passenger groups is considered. 3.2 The Synchronisation Quality Index (SQI) Fluerent et al (2004) developed the Synchronisation Quality Index as a eans of ore elaborately easuring the concerns of users in relation to schedule coordination. Two ain concerns were built into their approach: The need to differentiate between iniu, axiu and ideal transfer ties; and The need to differentiate the iportance of soe transfers over others. For exaple transfers onto CBD bound couter services can have a higher priority than counterpeak connections. Also soe interchange locations ight have ore network wide strategic significance copared to others. Figure 1 shows the basic coponents related to a transfer. Figure 1 : Key Eleents of a Transfer Trip (Fleurent et al, 2004) Passengers ake a bus rail transfer fro the on trip (bus) to the related trip (rail). The adissible wait interval is the tie where a transfer is possible or feasible. A transfer is feasible if a iniu wait tie is allowed such that passengers can walk between the on trip to the related trip. The adissible tie interval also has a axiu wait tie. This can be set as the tie which the slowest person ight require to ake the transfer. It ay also be set to ensure transfers are possible allowing soe eleent of leeway for unreliable services. Soe tie cushion ay also be required to avoid the next related trip based on the rail 28 th Australasian Transport Research Foru Page 4

5 service headway. An ideal wait tie ight lie soewhere between the iniu and axiu. This could be set near the iniu to ake for tight transfers or be close to the axiu to ensure ore people can achieve a transfer in unreliable conditions. Each interaction between an on trip and a related trip is tered a trip eet. In this case it is a feasible eet. If the transfer takes less than the iniu wait tie or ore than the axiu wait tie it is not an adissible wait interval. It is also tered an unfeasible eet. Fluerent et al (2004) present a forula to easure the quality of coordination tered the Synchronization Quality Index or SQI. The SQI is calculated for each trip eet using Forula 1: SQI QI QI = QI QI = w = w = 0, a = w ( Iin + (( a l )/( i l ))( Iax Iin )), a [ l, i] ( Iax + (( i a )/( u i ))( Iax Iin )), a [ i, u] [ l, u], H UnfeasCosta, [ l, u ], H, where: w = weight to apply to trip eet I in = iniu quality index base value for feasible eets I ax = axiu quality index base value for feasible eets a = actual wait tie of eet l = iniu wait tie for eet i = ideal wait tie for eet u = axiu wait tie for eet UnfeasCost = cost for historical feasible eets H = set of all historical eets. This forula follows the process identified in Figure 2. Forula 1 1 Is Actual Wait Between Ideal Wait? Yes QI = w ( I + ( ( a l ) ( i l ) )( I ) ) in / ax I in No Is Actual Wait Between Ideal Wait & Max Wait? Yes QI = w ( I + ( ( i a ) ( u i ) )( I ) ) ax / ax I in No Have Tietable Changes Made a Historical Meet Unfeasible? Yes QI = w UnfeasCost No QI = 0 Figure 2 : Process of SQI Measureent 1 Note: The second line in this equation has been altered fro the original for in Fleurent et al (2004). We have consulted with the authors who agree with this adjustent. In addition the for of the equation with respect to historical eets has been adjusted to add weighting for trip eets. 28 th Australasian Transport Research Foru Page 5

6 The SQI value is large and positive for trip eets where actual wait tie is equivalent to ideal wait tie i.e. where coordination is good. The forula for calculating SQI is different if actual wait tie is between the iniu and ideal wait than it is if it is between the ideal and axiu waits. In addition a separate process is used if a feasible eet becoes unfeasible as a result of changes in the schedule. This is tered a historical eet i.e. one that used to exist before the tietable change. A penalty is applied if a historical eet is no longer feasible. Tered UnfeasCost it is applied to create a negative value for SQI. Finally if a trip eet is unfeasible i.e. the actual wait tie is less than the iniu or ore than the axiu but has not changed as a result of changes in the tietable, then the SQI value is zero. In addition to the above, values for SQI are weighted (w ) to represent the level of iportance of the trip eet. Figure 3 shows how values for SQI vary under various scenarios of transfer tie. SQI w I ax w I in 0 Transfer Tie Unfeasible eets which were historically unfeasible Miniu wait Ideal wait Maxiu wait Unfeasible eets which were historically unfeasible w UnfeasCost Unfeasible eets which were historically feasible Figure 3 : Relationship Between SQI and Wait Tie Source: Based on Fleurent et al (2004) The highest (and best) values of SQI result where the actual wait tie is closest to the ideal. The index is zero if the trip eet is unfeasible and has always been unfeasible. However it is a negative nuber i.e. the value of the penalty UnfeasCost if the trip eet is unfeasible but used to be historically feasible before the change in the tietable. 4 Application of SQI The SQI approach was applied in a siple bus rail tietable case study in Melbourne on bus route 630. Coordination quality was assessed at 3 stations between route 630 and the 3 separate rail lines it crosses at Huntingdale, Orond and Gardenvale Railway Stations. At each station only transfers between bus route 630 and rail was considered. However this was reviewed in all directions. 28 th Australasian Transport Research Foru Page 6

7 Site investigation and a review of tietables suggested values for the paraeters as indicated in Table 1 for the a.. peak. Off peak values were adjusted for u (axiu wait tie) and UnfeasCost (penalty for reoving a feasible trip eet) because these paraeters are related to service headways. Table 1 : Paraeter Settings for SQI Application Melbourne Bus-Rail Case Study A.M. Peak Paraeter Value Basis a actual wait tie As indicated in Based on an analysis of schedules tietable l iniu wait tie 2 ins Based on site review of a iniu tie at a odest walk pace i ideal wait tie 5 ins Enables reasonable walk plus a odest tie on platfors u axiu wait tie 10 ins Ideal wait tie plus 6 ins which is equivalent to around 95% of late running range for rail and bus At values over 10ins passengers would see UnfeasCos t I ax I in w Penalty for historical feasible eets which are no unfeasible Maxiu quality index base value for feasible eets iniu quality index base value for feasible eets weight to apply to trip eet leaving trains as the bus arrives -20 ins Based on the wait tie which passengers would have to wait until the next train in the headway (10 ins) Weighted by passenger valuations of wait and walk tie. A weight of 2.0 is used based on Transfund New Zealand (2000) 10 5 This is the scaling factor which deterines the relative size of index values. A siple scaling between 5 and 10 was used based on its siplicity 10 High weight for CBD bound trains in the a.. peak 5 Lower weight for counter peak connections Figure 4 shows the resulting total values of SQI for each of the three stations for a weekday, Saturday and Sunday. 20,000 Total Station SQI Values 18,000 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 Huntingdale Orond Gardenvale Weekday Saturday Sunday Figure 4 : Station SQI Values Day of Week 28 th Australasian Transport Research Foru Page 7

8 This suggests that weekday coordination of service is better than at weekends. This ight be expected because weekday headways are shorter than at weekends. Hence there is ore opportunity to transfer within a reasonably low tie. Figure 4 also suggests that Huntingdale has better coordination than Orond or Gardenvale Station on weekdays. Also that Gardenvale has better coordination than Orond on weekdays. Closer analysis of the SQI coputations in this case suggests that these conclusions are not quite so clear. Table 2 shows the nuber of vehicle trips run by bus and rail at each of the stations, the nuber of feasible trip eets and the SQI values coputed. Table 2 : Schedule Coordination Inputs and SQI Outputs Melbourne Bus-Rail Case Study Station No. Vehicles Trip/ Day Feasible Trip Meets SQI Rail Bus Weekday Huntingdale ,170 Orond ,902 Gardenvale ,446 Saturday Huntingdale ,650 Orond ,662 Gardenvale ,791 Sunday Huntingdale ,448 Orond ,120 Gardenvale ,310 The weekly distribution of coordination quality is driven by the nuber of vehicle trips where transfers can occur as uch as it is by coordination quality. Hence Huntingdale Station, which has 156/152 rail/bus vehicle trips each weekday will have a uch higher SQI score because there are ore trips copared to Gardenvale on a Sunday with 94/40. While certainly ore feasible trip eets are ade in the Huntingdale case, the nuber of vehicle trips ade at each station biases the results towards a high SQI outcoe as uch as coordination quality does. The iportant point here is that it is possible for a very large station with any coordinating services but with poor quality of trip eets (long waiting ties) to have a higher total SQI score copared to a station with only one coordinating bus at very high quality. The SQI index is easuring ore than just coordination quality. It is also easuring scale of service which is not directly an ai of the easure. Another iportant point is that the SQI scores theselves don t ean uch. When Huntingdale has a total SQI of 19,170 (weekday) and Gardenvale 2,310 (Sunday) it is difficult to say uch based on these values other than they are a long way apart. Because of these issues the research project explored ways in which SQI ight be adjusted to create a ore eaningful easure. 5 Developent of the Synchronisation Quality Ratio The developent of an iproved SQI easureent sought to reove the bias which high volues of trip eets akes to the output index. A Synchronisation Quality Ratio (SQR) approach was developed which identified the axiu possible SQI which could be 28 th Australasian Transport Research Foru Page 8

9 coputed. This includes any trip eet weighting and would assue all eets are feasible and waiting ties are ideal. The SQR is thus calculated as the proportion of actual SQI in relation to this potential axiu as shown in Forula 2. SQI SQR = Forula 2 SQI Max The SQI ratio thus returns a ratio or percentage which is representative of the quality of coordination relative to the axiu possible quality which could ever be achieved i.e. every bus eets every train at the ideal waiting tie. While SQI ax ay never be achievable on a network wide basis, it is still an interesting nuber to be aware of, since it sets a standard which is high. Figure 5 shows how the results fro the SQR copare with the absolute values in the SQI forula for the stations in the Melbourne bus-rail exaple. SQI % Synchronization Quality Index Synchronization Quality Ratio 60% 50% 50% 40% 40% 30% SQR 20% 20% 10% 10% 0% Weekday Saturday0% Sunday Weekday Saturday Sunday Weekday Saturday Sunday 30% Huntingdale Orond Gardenvale Figure 5 : Coparison of SQI and SQR Measures of Schedule Coordination The two approaches suggest rather different values for coordination quality at each station by tie period. SQR acts to bring together weekend and weekday valuations although the poorer relative quality of weekend day to weekday is retained, it s just saller. The sae is true about the relative gaps in coordination quality between stations. Interestingly the rank of station coordination quality on Saturdays changes according to the ethod used. The best SQI for Saturday coordination is at Gardenvale while SQR has Huntingdale with the best coordination quality. Overall SQR was considered a better easure of relative coordination quality therefore was adopted for further investigation. 6 Exploring the Synchronisation Quality Ratio A series of tests were undertaken to explore how the SQR would perfor in ters of sensitivity to changes in key paraeters and variation in relation to the adjustent of tietables. Sensitivity tests exained variation in iniu wait tie (l ), axiu wait tie (u ) and ideal wait tie (i ). A test of tietable adjustents was also undertaken for peak bus services. 28 th Australasian Transport Research Foru Page 9

10 6.1 Variation in Miniu Wait Tie (l ) Figure 6 shows the ipacts of increasing iniu wait tie on the SQR at each station. 58.0% Huntingdale Orond Gardenvale 56.0% 54.0% 52.0% 50.0% SQR 48.0% 46.0% 44.0% 42.0% 40.0% 38.0% 0:01 0:02 0:03 0:04 Miniu Wait Tie (l ) Figure 6 : Variation in Miniu Wait Tie (l ) and SQR Increasing iniu wait tie acts to reduce SQR because ore trip eets becoe unfeasible. Stations are ipacted differently depending on the distribution of transfer ties. Orond is less ipacted by an increase in l fro 1 inute to 2 inutes than Gardenvale is. This is because there are less transfer eets affected at Orond. 6.2 Variation in Maxiu Wait Tie (u ) Figure 7 shows the ipacts of variation in axiu wait tie on the SQR at each station. 75.0% Huntingdale Orond Gardenvale 70.0% 65.0% 60.0% SQR 55.0% 50.0% 45.0% 40.0% 35.0% 0:08 0:09 0:10 0:11 0:12 0:13 0:14 0:15 Maxiu Wait Tie (u ) Figure 7 : Variation in Maxiu Wait Tie (u ) and SQR As axiu wait tie increases ore trip eets becoe feasible and SQR increases as a result. The rate of increase flattens out at higher axiu wait ties because there is only a finite nuber of feasible trip eets so the benefit of getting additional eets to be feasible 28 th Australasian Transport Research Foru Page 10

11 reduces. Even when all trip eets are feasible SQR is not 100% because this can only happen when the actual waiting ties are equal to the ideal. Figure 7 shows that stations are affected differently by variation in axiu wait paraeters. This is because the nuber of trip eets which are not feasible is different at each station. 6.3 Variation in Ideal Wait Tie (i ) Figure 8 shows the ipacts of variation in ideal wait tie on the SQR at each station. 53.5% Huntingdale Orond Gardenvale 53.0% 52.5% 52.0% SQR 51.5% 51.0% 50.5% 50.0% 49.5% 0:04 0:05 0:06 0:07 Ideal Wait Tie (i ) Figure 8 : Variation in Ideal Wait Tie (i ) and SQR In general, Orond and Gardenvale have better perforance when ideal wait ties are set at 5 ins while Huntingdale has an optial ideal tie at 4 inutes. This illustrates that transfer ties at different stations are distributed in different ways. 6.4 Ipacts of Adjusting Peak Bus Schedules A separate set of SQR odels were set up to odel trip eets for these stations in the a.. peak. Soe 8 sets of bus vehicle trips were odelled (trips A to H) and a trip shifting function added such that shifting the tiings of an individual vehicle trip adjusted trip eet ties at each station. The odel coputed SQR values for adjusting each of the 8 vehicle trips separately. A separate odel was used to copute total SQR values for cobinations of trip shifts. Figure 9 shows the outputs fro the individual trip shifting odel. Higher values for SQR are achieved by : Shifting the bus G tietable by 1 inute Shifting the bus C tietable by +2 inutes. 28 th Australasian Transport Research Foru Page 11

12 Iproved Coordination 53.0% G D D SQR for Each Tietable Shift G 52.0% D 51.0% E 50.0% C 49.0% H 48.0% C D H E A B F C A E B G H C A E F B 47.0% B A Bus A Bus B Bus C Bus D Bus E Bus F Bus G Bus H F G H 46.0% F Tietable Shift ( in inutes) Figure 9 : Variation in Maxiu Wait Tie (u ) and SQR Tests indicated that by cobining these two adjustents the SQR could be increased to 53.2% an overall increase of about 2% in total fro the base tietable. Although this benefit is not to be ignored it is hardly substantial. The sall scale of this benefit and the larger range of negative ipacts associated with other tietable shifts suggests that the base tietable is relatively robust and well coordinated. 7 Conclusions This paper has reviewed a series of issues associated with schedule coordination. Fro a user perspective good schedule coordination is a balance between allowing enough tie to coplete a transfer against iniising the wait tie between connections. In designing schedules any other factors including tangible resource cost iplications ust be traded off against factors such as transfer quality. The coplexity of transfer coordination is not well understood by users who express concerns over tangible localised isatches at local interchanges. Planners understand the issue of coordination but lack effective tools to easure coordination quality. Most planning for schedule coordination is intuitive as a result. The use of syste wide network schedule coordination software to optiise bus, rail (and tra) schedule coordination is not a feasible option for ost cities. In practice separate operators have different approaches to scheduling and solutions for schedule coordination will require a tool which can copare schedules between transit odes, services and separate operators in order to understand and easure the overall quality of coordination. Without an open and defendable basis for easuring the proble, user coplaints about coordination will always be difficult to retort. Planners are open to uch criticis because they have no technical defence. Measures of coordination quality were reviewed. Use of passenger wait tie and the SQI approach were shown to have weaknesses. A new approach tered the Synchronisation 28 th Australasian Transport Research Foru Page 12

13 Quality Ratio was developed and tested under a range of sensitivity tests. It has a nuber of benefits in that: It only requires tietable data which is generally readily available as a result of the increased digitisation of web based schedule inforation systes. It can be used to easure coordination quality for ulti-operator situations. There is an opportunity to develop a eta-syste where schedules fro any sources can be cobined and assessed to deterine overall coordination quality. It is sensitive to coordination quality and not the volue of services. It can be weighted to highlight iportant transfer connections and be representative of passenger volue or scale of interchange iportance. SQR requires calibration to identify appropriate paraeter setting for individual interchange sites. A range of potential settings were explored by sensitivity testing in a sall case study. A nuber of steps are required to further develop the approach. There are also a nuber of research questions to be explored: A wider exploration of paraeter sensitivity is warranted. SQR needs to be tested on a larger data set. Its ipleentation on a city wide basis will require software developent and standardisation of schedule data forats. A coputational challenge is that ost tietables only show arrival ties at a few points 3-5 on ost routes. In addition they often do not indicate dwell ties, ties indicted are averaged and are often not accurate. SQR needs a good estiate of arrival and departure ties for all locations where coordination is possible. The increasing developent of tietable inforation tools such as trip planners has ade schedule data of this type ore readily available. The researchers have already had discussions with MetLink in Melbourne exploring the availability of such data. The developent of a coputer package which easures coordination quality and can easily assess tietables of any forats would be an obvious long ter objective. Such a tool could be developed to highlight proble locations and even undertake a preliinary assessent of where better coordination ight be feasibly achieved. Use of large scale scheduling systes to undertake such a task would be expensive but is possible. A saller scale syste to quickly identify opportunities which individual operators ay then go on to explore and ipleent using their own tools would be a ore pragatic and cost effective approach to the proble. It would be beneficial to explore the relationship between passenger wait tie easures of schedule coordination with SQR. Passenger wait tie easures have the benefit of using the volue of users as a easure. This is a better easure than the weighted optiisation applied in SQR however SQR reains a ore practical way of tackling the proble. Exploring ways to calibrate SQR so it is a better representation of passenger wait tie will be an effective advance in the approach. Acknowledgents The authors would like to thank Professor Avi Ceder and Mr Charles Fleurent for coents and inputs to this work. Any oissions or errors are the responsibility of the authors. 28 th Australasian Transport Research Foru Page 13

14 References Ceder, A Golany, B and Tal, O (2001), Creating Bus Tietables with Maxial Synchronization, Transportation Research, Vol. 35A, No. 10, pp , Fleurent C., Lessard R. and Seguin L. Transit Tietable Synchronization: Evaluation and Optiization 9th International Conference on Coputer Aided Scheduling in Public Transport. San Diego California August Rousseau, JM and Haer, N (1993) Case Studies in Transit Schedule Optiization Aerican Public Transport Association Interodal Planning Conference 1993 Transfund New Zealand (2000) Valuation of Public Transport Attributes Booz Allen Hailton Final Report to Transfund New Zealand Research Prograe Voss S (1992) Network Design Forulation in Schedule Synchronization In Coputer- Aided Transit Scheduling, Desrochers and Roussea (Eds.) Springer Verlag Berlin- Hiedelberg 1992, pp Wong R.C.W. & Leung J.M.Y. Tietable Synchronization for Mass Transit Railway 9th International Conference on Coputer Aided Scheduling in Public Transport. San Diego California August Yeates, M (2002) Transport challenges still looing unanswered The Brisbane Institute 28 th Australasian Transport Research Foru Page 14