OMDS/BALOR Talk Series

Upcoming Speaker: Laura Kolcheva (Polytechnique Montréal)

 

The OMDS/BALOR Talk Series invites you to a seminar talk in September 2025: 

 

  • Speaker: Laura Kolcheva (Polytechnique Montréal)
  • Title: Real-time transfer synchronization for public transit networks
  • Time: Thursday, Sept. 25, 2025, 14:00 - 15:00
  • Location: OMDS Seminar Room (03.307), Oskar-Morgenstern-Platz 1
  • Abstract: Public transit plays a crucial role in urban development and in reducing congestion and greenhouse gas emissions. The reliability and timeliness of transfers are key factors influencing ridership and user experience in public transit networks. Real-time transfer synchronization in public transportation networks is an essential approach for improving service quality and reducing passenger travel times. We develop and evaluate dynamic optimization approaches that enhance transfer reliability by implementing real-time control tactics such as holding, skip-stop, and speed control.
    • First, an offline arc-flow model is developed using time-expanded graphs to integrate all possible control tactics within a given control horizon. The model minimizes total passenger travel time by optimizing transfers while reducing schedule deviations. As a proof of concept, it demonstrates the ability to handle large-scale instances, involving several dozen control stops and numerous transfer connections, in real time. To assess the performance of the offline model in an environment with partial information, a dynamic version generating a single scenario is designed. Different levels of uncertainty are evaluated in the scenario generation. Both methodologies are tested using real-world passenger demand and ridership data from bus route 70 of the Société de Transport de Laval (STL) network. Results confirm the effectiveness of this approach for transfer synchronization and indicate that, when considering a single scenario, a deterministic scenario generation method yields the best performance. However, a performance gap remains between the perfect information offline model and its deterministic dynamic counterpart.
    • Next, a discrete-event simulation framework is developed to incorporate the inherent uncertainties of traffic conditions and passenger flows. Two online stochastic optimization (OSO) algorithms, Consensus and Regret, are adapted from the literature for the transfer synchronization problem, leveraging historical and real-time data to generate multiple scenarios and make robust decisions. These algorithms evaluate the solutions of the offline model for all scenarios to determine optimal control tactics at stops. Tested on 29 bus routes within the STL network, the algorithms significantly improve successful transfer rates and reduce total passenger travel time. The results demonstrate that OSO achieves performance levels comparable to the offline model while maintaining adaptability to dynamic network conditions.
    • Finally, the approach is extended to the full STL network using a network-wide simulator replicating real-time stochastic conditions based on comprehensive historical data on network operations and passenger demand. The offline model and the online algorithms are adapted and implemented in the simulator. The objective is to evaluate the impact of the simultaneous optimization of multiple routes on overall network performance. The study also examines various contexts, including high- and low-frequency routes as well as grid and radial sub-networks. This analysis highlights the potential of network-wide synchronization to improve transfer reliability without causing excessive disruptions. These methodologies, which combine computational efficiency with robustness, open concrete application opportunities for transport operators.