An Adaptive Traffic Vehicle Routing Framework During Big Public Events
MetadataShow full item record
One of the most ubiquitous transportation problems is real-time vehicle routing. Previous work leveraged both real-time (e.g. vehicle location and velocity) and historical traffic data (e.g. street traffic count within a certain amount of time) collected from traffic sensors and road cameras, to facilitate vehicle routing. However, one type of data, vehicle travel plans, were commonly ignored. Travel plans are the data which describe: 1) the origin a vehicle is from; 2) the destination it is heading to; and 3) the time it plans to leave from the origin. The travel plan data is valuable for vehicle routing because current travel plans from vehicles could cause corresponding traffic congestion somewhere in the network sometime in near future, which is defined as “butterfly effect” in this thesis. Therefore, we could improve future travel plans by avoiding the anticipated traffic congestion spots predicted by these known travel plans. Due to the fact that the state-of-art work tends to route each vehicle independently without considering such “butterfly effect” among vehicles, this paper proposed an adaptive traffic routing framework which would fully use traffic plan data to improve real-time routing and mitigate traffic congestion. Additionally, the state-of-the-art work failed to make a comprehensive study of vehicle routing during big public events (e.g. a football game in a stadium or a movie in a theater), where there are two phases, including the “black hole” phase caused by a large number of vehicles heading to the same destination before the public event begins, and the “volcano” phase when vehicles leave from the event spot and drive to different directions after the event. While previous work may focus on addressing either the black hole phase or the volcano phase, none have attempted to study the two phases together. Therefore, this paper develops a new framework which can handle both phases at the same time. To perform vehicle routing, traditional approaches widely use A* algorithm, which didn’t consider the influence of the travel plans (i.e., the butterfly effect). Therefore, a time dependent adaptive traffic routing algorithm, which consistently updates the traffic conditions with current travel plans, is implemented and integrated in a traffic simulation system, known as MATSim, within the proposed framework. To demonstrate the feasibility and effectiveness of this adaptive traffic routing framework, a series of experiments are conducted based on the MATSim system. The experiment results show that our framework is able to leverage the butterfly effect among vehicles for vehicle routing. The framework can also handle a large number of vehicles for both the black hole phase and the volcano phase. Our framework outperformed the traditional A* routing algorithm in both phases, with approximately 15% and 45% performance improvement, respectively, in all experiments.
Adaptive traffic routing
Includes table of contents, figures, maps, tables, equations, images, bibliography