A methodology for the simulation based assessment of coordination strategies for signalized networks using Particle Swarm Optimization

Abstract

The coordination of signal programs at adjacent intersections offers many advantages, the most important being that travel times and emissions can be reduced. However, coordination is a multivariate optimization problem with many constraints. Numerous optimization strategies have been devised in the past. The quality of these coordination strategies depends on three major factors: the quality of the optimization procedure itself, the performance measures it is based upon, and the quality of the input data. In real world applications, the impact of the latter factor in relation to the other factors is commonly a source of great uncertainty. To be able to analyze all three factors, a methodology has been developed to compare different offline optimization strategies using complete information of the traffic flow. The complete information is obtained by using a mesoscopic traffic flow simulation (AVENUE). To have a benchmark for a given optimization function, the best possible coordination for given conditions (traffic demand etc.) is computed using Particle Swarm Optimization (PSO). Different strategies can then be compared to this benchmark and with each other. Furthermore, the stability of the strategies under changing traffic demand and the effect on different performance indices (e.g. number of stops, average delay) can be evaluated. This article describes the methodology, expands upon Particle Swarm Optimiza-tion as a useful tool in signal control, and highlights the opportunities arising out of the chosen approach.