Multi Agent Optimizing Traffic Lights using Reinforcement Learning

B Sai Siri1, P L Keerthana Reddy2, G Bensun3, E Venu4, Mrs L Lavanya 5
1 Department of Computer Science and Engineering(AI&ML), Student, Sri Venkateswara College of Engineering
2Department of Computer Science and Engineering(AI&ML), Student, Sri Venkateswara College of Engineering
3Department of Computer Science and Engineering(AI&ML), Student, Sri Venkateswara College of Engineering
4Department of Computer Science and Engineering(AI&ML), Student, Sri Venkateswara College of Engineering
5Department of Computer Science and Engineering(AI&ML), Assistant Professor, Sri Venkateswara College of Engineering

Abstract--- Urban traffic congestion has escalated due to rapid vehicle population growth and the inability of conventional fixed-time signal systems to adapt to dynamic traffic conditions, resulting in increased delays, fuel wastage, and pollution. This project proposes an intelligent traffic signal control system based on Deep Reinforcement Learning (DRL), utilizing a Multi-Agent framework where each intersection is independently controlled by a learning agent. These agents analyze real time parameters such as vehicle density, queue length, and waiting time to dynamically optimize signal phases. The system is developed and evaluated using the SUMO traffic simulator, with the Deep Q- etwork (DQN) algorithm enabling agents to learn optimal control strategies. An emergency vehicle prioritization mechanism is further incorporated to ensure unimpeded passage during critical situations. Experimental results demonstrate notable reductions in congestion and average waiting time, confirming the system's suitability for smart citydeployments.