Distributed Architecture for Real-Time Coordination in Transit Networks

Transit is one of the vital service sectors of the present and the future US economy, and it holds tremendous social significance as an estimated 25% US citizens rely on some form of a transit system. Owing to their inherent flexibility and low capital costs, bus transit system are most attractive to meeting the growing transportation requirements. Despite their vital importance, conventional bus operations are perceived in general to be unreliable. Inadequate service reliability is manifested as large waiting times for passengers. In order to overcome these problems, modern transit systems are beginning to be equipped with Intelligent Transportation Systems (ITS), such as Global Positioning Systems (GPS) and Mobile Data Terminals, in order to provide real-time information to passengers as well as dispatchers. Real-time coordination of how various buses are dispatched and how long the buses are held at various stations on a transit network, harnessing the information from these ITS, is known to be essential for substantially increasing operational efficiency, in addition to reducing the passenger waiting time and operational delays In this project, we have developed a negotiation-based framework to address the real-time coordination of bus holding, wherein stops and buses act as agents that communicate in real-time to achieve dynamic coordination of bus dispatching at various stops. They negotiate based on their marginal costs. Our work proves that under some reasonable assumptions, the framework can find a local optimal dispatching time. The algorithm is further modified to overcome the “myopia” of the local optimality. To verify the efficiency of our framework, we have compared our framework with other simple bus control strategies such as On-Schedule and Even-Headway strategies, through simulations. In order to show the robustness of our framework, we have tested it under several representative transit environments. The simulation results show that our framework distinguishes itself for its ability to harness real-time information and to make decisions directly based on the marginal wait cost. Furthermore, the framework has been found to be efficient in terms of minimizing passenger wait costs, and adequately robust in order to be applicable to a wide range of transit environments involving both stationary passenger arrivals as well as a variety of non-stationary passenger arrivals, including those with random bursts.