A Comparison of Tram Priority at Signalized Intersections in Melbourne Lele Zhang, Timothy M. Garoni and Somayeh Shiri School of Mathematical Science, Monash University, Victoria 3800, Australia

Introduction Simulation Results Person Performance Figs. 2 (g-h) give the aver- Transit priority at signalized intersections is an ef- observables age person travel time and throughput of the whole fective strategy to improve tram performance. Com- - Throughput Total number of cars (trams) that network as the number of occupants that each tram PC pared to buses, trams operating in mixed traffic have have traversed the network during the simulation carries (0t,p), varies. provides the shortest av- much higher impact on other road users, and vice time. erage person travel time in both US and OS. Ab- versa. - Aggregated travel time per car (tram) The ex- solute tram priority provides shorter person travel time when tram occupancy is sufficiently high. • pected value of travel time over all traversed cars( Trams block the entire link when they stop. Figure 1: Left. Illustration of the studied network with each • Trams cannot change lane. alternating east-west route being a tram route. All links carry tram) during the simulation time. Conclusion - Travel time variability The variance of the aggre- • It is comparatively difficult for early trams to bidirectional traffic. Boundary links are treated as ramps (buffer- ing zones) for inputting and outputting vehicles. Right: Signal Tram priority improves tram performance in terms stay on schedule. gated travel time per car (tram). phases. Dashed paths are required to give way to continuous of both travel time and variability. Regardless of the

OS OS OS 45 150 18 paths. Phases E and F are for tram nodes only. W−to−E E−to−W We study the effect of different levels of tram priority 40 16 traffic condition, the absolute tram priority results in 35 14 (min) t

30 100 σ 12 currently used, or being considered, in Melbourne, Boundary conditions There are two peak direc- 25 10 the best tram service in the priority direction at the 20 8 tram throughput Australia, by simulations on large-scale networks tram travel time (min) 15 50 6 expense of delaying other traffic in the non-priority tions: eastbound and southbound. The inflow rates 10 tram travel time deviation 4

5 2 W−to−E E−to−W W−to−E E−to−W 0 0 0 governed by realistic adaptive signal systems with NT PU PC AU AC NT PU PC AU AC NT PU PC AU AC in the peak directions are about twice as large as signal scheme signal scheme signal scheme directions. open boundary conditions. those in the counter-peak directions during the peak (a) Mean tram travel time (OS) (b) Mean tram throughput for all 4 routes (OS) (c) Tram travel time variability (OS) As the network becomes more saturated however,

Tram route: OS OS Non−tram route: OS hours. The boundary condition follows an AM pro- 40 25 12 other road users suffer more from the disruptions

35 10 Transit priority variants 20 30 caused by tram priority processes, especially the ab- file. We consider two scenarios: over-saturated (OS) 8 25 15

20 6

10 solute tram priority. and unsaturated (US). In the OS scenario the net- 15 - Absolute priority starts the priority phase im- 4 10 private vehicle travel time (min) private vehicle travel time (min) private vehicle travel time (min) 5 2 5 For either the absolute priority or the partial pri- work density reaches higher than US. W−to−E E−to−W N−to−S S−to−N W−to−E E−to−W mediately after detecting a tram and keep a pri- 0 0 0 NT PU PC AU AC NT PU PC AU AC NT PU PC AU AC signal scheme signal scheme signal scheme ority, the conditional version achieves almost the ority phase running until the tram traverses the Turing decisions Each link was assigned with a (d) W-E tram routes (OS) (e) N-S routes (OS) (f) W-E non-tram routes (OS) same level of improvement of service as the uncon- OS, fixed o /o = 4 OS, fixed o /o = 4 . . t,p t,n 5 t,p t,n OS probability 0 9 of continuing straight ahead, 0 04 of x 10 approaching intersection. 19.5 1.6 14

19 1.55 12 ditional version but with reduced impact on other 18.5

turning into a non-peak direction link, and a prob- 1.5 (min) p 10 - Partial priority has less disruptive priority tac- σ 18 1.45 8 17.5 traffic. Therefore, the partial conditional priority ability 0.06 of turning into a peak-direction link. 1.4 6 17

tics, which include a clearance phase and a green person throughput person travel time (min) 1.35 4 16.5

person travel time deviation system appears worth trialling. In the case that the 16 1.3 2 extension and operates up to 20% of the cycle NT PU PC AU AC NT PU PC AU AC Traffic signal systems 15.5 1.25 0 40 60 80 100 120 140 160 180 200 40 60 80 100 120 140 160 180 200 NT PU PC AU AC o o t,p t,p signal scheme absolute tram priority is necessary, the absolute con- time. (g) Mean person travel time (OS) (h) Mean person throughput (OS) (i) Person travel time variability (OS) - Conditional priority is active only when trams SCATS signal coordination (linking) was set in the ditional priority should be implemented. are behind schedule. eastbound direction, to establish green-wave behav- Figure 2: performance of versions of SCATS. Error bars corre- sponding to one standard deviation are shown but are usually References ior. Cycle length and split plan are adaptive, i.e. too small to observe Unconditional partial priority is the system cur- more congested, more green time [2]. Tram Performance [1] J. de Gier, T. Garoni, O. Rojas, Journal of Statistical rently employed in Melbourne. Figs. 2(a-c) shows tram per- Version Description formance in OS case. In terms of mean travel time Mechanics: Theory and Experiment 4 (2011) NT Multimodal Traffic Model SCATS with no tram priority. Tram priority reduces the average tram travel time [2] L. Zhang, T.M. Garoni, J. de Gier, Transportation PU SCATS with partial and unconditional tram priority. in the eastbound direction, when compared to the Research Part B: Methodological 49, 1 (2013) We utilize a multimodal CA model extends the Net- PC SCATS with partial and conditional tram priority. AU SCATS with absolute and unconditional tram priority. no priority system NT, and the improvement is more NaSch unimodal traffic model, ([1]) on a generic 8 by AC SCATS with absolute and conditional tram priority. significant under the OS scenario than US. Acknowledgements 8 square-lattice, as shown in Fig. 1 (left), governed Table 1: Versions of SCATS Private Vehicle Performance Figs. 2(d-f) by SCATS (Sydney Coordinated Adaptive Traffic We gratefully acknowledge the financial support of the Roads - NT assigns 20% of the cycle length to either phase shows mean car travel times along different ap- Corporation of Victoria (VicRoads). This work was supported System). Each link is 750m long. Each tram links E or F (only tram nodes phases). proaches in OS case in different routs. When tram under the Australian Research Council’s Linkage Projects has two lanes: left car lane and right mixed-traffic - When the tram has passed the middle-link detec- priority process is active, both southbound and funding scheme (project number LP120100258). This research lane, three tram stops and two tram detectors. A was undertaken with the assistance of resources provided at the tor, PU runs right-turning clearance phase to clear northbound travel times increase. The higher the non-tram link has two car lanes plus a right-turn NCI National Computational Merit Allocation Scheme sup- a passage ahead of the tram. When the tram has priority imposed, the more the north-south traffic lane of 90m. ported by the Australian Government. passed the end-link detector but has not traversed gets penalized. Interestingly, from Fig. 2(f) tram Each private car occupies one cell of 7.5m with max- the intersection, PU runs extension phase. priority can penalize the traffic in parallel non-tram Contact Information imum speed of 3, v = 3, and each tram occupies max - When the tram priority process is triggered, AU routes. Perhaps surprisingly, the penalty generated 3 cells, i.e. 22.5m long with v = 2 as trams travel max starts the right-turning clearance phase and keeps by PU and PC is larger than that by AU and AC. • [email protected] more slowly than private cars in mixed traffic. running it until the tram has traversed the intersec- • [email protected] ————————————————————— tion. • [email protected] ————————