Agent-based intelligent traffic signal control at urban intersections

Abstract

Increased stress, air pollution and delays are among the effects of traffic jams in urban areas. Therefore, the objective of our work is to minimize vehicle’s waiting time, CO2 emissions and fuel consumption, in order to improve traffic flow. This work is situated in the context of traffic management and intelligent transportation. Based on reinforcement learning, we have realized an intelligent traffic signal control system for urban intersections (SICoFSv2), built on the multi-agent approach using JADE (Java Agent Development Framework) as a tool to facilitate the implementation, and SUMO (Simulation of Urban MObility) as a simulator for our system. We proceeded with the experimentation in order to compare our system with others, for a good improvement of the performance. In the end, we succeeded in reducing vehicle waiting times, CO2 emissions and fuel consumption by a very high rate.