Experiments and Evaluation of an Adaptive Traffic Light Control Method Based on a Multi-Agent Model

Abstract

Traffic congestion has become chronic in Japan. Reducing traffic congestion is very important not only to reduce time loss for drivers, but also to curb exhaust emissions that contribute to global warming. We propose a model for traffic light control based on real-time traffic information that combines split control and offset control, aiming at smoothing traffic. We further implemented this method on traffic signals located on real roads and conducted experiments with real traffic flows. In the initial stage of the experiment, the proposed control was found to be effective to a certain extent; however, there were cases where effective control could not be achieved due to factors such as the number of vehicles passing through. Therefore, we added a new control related to cycle length, adjusted the parameters, and conducted the demonstration experiment again. This paper outlines the implementation model and evaluates its performance.