Maximum-stability Distributed Control in Traffic Networks

Abstract

The max-pressure control is a distributed control algorithm that has the property of stabilizing the total queue length in the network theoretically. In spite of its good properties, some assumptions or requirements of the max-pressure control make it hard to be applied to traffic networks in reality: such as the data collection of queue length information for each movement and fixed route choices. Besides, traditional max-pressure control algorithms are only designed for signal-controlled intersections and are not applicable for signal-free intersections. Therefore, this thesis proposes max-pressure control algorithms and tests their performances in traffic networks while relaxing some of the assumptions used in existing studies. This thesis first explores mild assumptions for weight functions to incorporate alternative data sources in max-pressure control. This thesis also proposes an autonomous intersection management (AIM) algorithm considering pedestrians using the max-pressure control. Besides, the performance of max-pressure control is tested when road users' route choice is considered using dynamic traffic assignment, and a routing guidance algorithm is also developed to modify road users' route choices and to improve network efficiency.