Strategies for on-line management of a Multi-Layered Public Transit System

Abstract

but will be provided in the full paper. We use Benders decomposition to solve the problem. Given the head routes, cabin paths are dynamically modified throughout the operations based on the destinations of on board passengers. The decision to join a convoy is based on the position, route and available capacity of the head, the elapsed time of on board passengers and the number of cabins at stations. The assignment of passengers to cabins is done as soon as possible after their arrival. Given the position of cabins and occupancies, passengers are assigned to the cabins with the earliest estimated arrival times to their destinations. Passengers who wait too long at their origin are assumed to opt-out.

SIMULATION AND RESULTS The proposed management framework has been embedded in a purpose-built simulation that replicates heads, cabins and passengers movements. We measure the performance of the system by the mean passenger delay time (as compared to direct taxi service) and by the opt-out fraction. The simulation is run on a case study network of 20 stations to be implemented in a real-world area of approximately 3 km2 in an Asian city. Interstation distance varies from 200 to 800 meters. The daily demand is in average 2400 passengers. We tested the system performance for several combinations of number of cabins (10-90), number of heads (20,40,60) and different head route plans. Specifically, we examined two head route plans designed by the operator (R2,R3) and the route plans resulting from the head routing optimization model (Opt). In addition, in terms of station capacities, 60 platforms in total are distributed in the system’s stations. For any tested configuration, output values are averaged over 100 different demand realizations generated using a Poisson process based on hourly demand rates. Initial results (excluded from this abstract) have demonstrated that empty cabin relocation strategies must be applied to prevent significant deterioration of the system performance. In Figure 1, we present the delay time and opt-out fraction for the various combinations described above. As can be observed, the marginal effect of adding more heads in the system is positive but decreasing. As for cabin fleet size, the best results are obtained when the number of cabins is close to total platform capacity. In terms of head routing, the configuration resulting from the formulated head route-planning problem has shown to outperform operator-designed configurations, independently of the other system characteristics. In the best system settings, mean passenger delay time is less than 5 minutes while optout fraction is about 5%.

Figure 1: Performance measures obtained through simulation for various cabin, heads and route configurations

CONCLUSION AND ON-GOING WORK In this work, we introduce the MLPTS, an innovative and complex public transportation system. We propose a method to manage real-time operations of the system. Through simulation, we demonstrate that proper configuration of the system enables handling the estimated demands very well. On-going work focuses on enhancing each of the decision modules by extending the information taken into account in the decision process. Specifically, better estimating the future state of the system will allow to plan and operate more efficiently. The final goal of this study is to build a coherent and effective framework for the on-line management of MLPTS. REFERENCES G. Desaulniers and M. Hickman. Public transit. Handbooks in Operations Research and Management Science, 14:69 - 127, 2007.

J.-F. Cordeau, P. Toth, and D. Vigo. A survey of optimization models for train routing and scheduling. Transportation Science, 32(4):380 - 404, 1998.

I. Andréasson. Vehicle Distribution in Large Personal Rapid Transit Systems. Transportation Research Record, 1451:95 - 99, 1994.

V. Pillac, M. Gendreau, C. Guéret, and A. L. Medaglia. A review of dynamic vehicle routing problems. European Journal of Operational Research, 225:1 - 11, 2013.

G. Berbeglia, J.-F. Cordeau, and G. Laporte. Dynamic pickup and delivery problems. European journal of operational research, 202(1):8 - 15, 2010.

P. Friese and J. Rambau. Online-optimization of multi-elevator transport systems with reoptimization algorithms based on set-partitioning models. Discrete Applied Mathematics, 154:1908 - 1931, 2006.