Intelligent Traffic Congestion and Collision Avoidance Using Multi-agent System Based on Reinforcement Learning

The number of vehicles in Palestine has significantly increased over the past decade leading to significant traffic congestion in cities. The narrow structure of roads within cities, coupled with a lack of development and updates, has exacerbated this problem. Congestion causes air pollution and driver frustration and costs a significant amount in fuel consumption. Additionally, collisions between vehicles waiting at traffic lights can occur due to high speeds or small distances between waiting cars. Finding solutions for this dynamic and unpredictable problem is a significant challenge. One proposed solution is to control traffic lights and redirect vehicles from congested roads to less crowded ones. A multi-agent system is utilized in this study. Based on the JaCaMo platform was developed to address the issue of traffic congestion and collision avoidance. Simulation using SUMO and JADE platforms demonstrated that traffic congestion could be reduced by 52.7% through traffic light timing control.