TrafficLightRL

Abstract

Addressing urban traffic congestion is crucial for environmental sustainability, as inefficient traffic flow leads to increased fuel consumption and greenhouse gas emissions. This paper presents TrafficLightRL, a reinforcement learning (RL)- based traffic light control system designed to minimize vehicle emissions and improve traffic efficiency. Using SUMO for simulation and the Proximal Policy Optimization (PPO) algorithm from Stable-Baselines3 for RL training, our system dynamically adapts to real-time traffic conditions. Results show that the RL agent reduces CO2 emissions by up to 11.6% compared to traditional fixed-time systems, with performance evaluated across various traffic densities. This study highlights the potential of RL-driven solutions to enhance traffic management and reduce environmental impact. The code and resources for this project are available at: https://github.com/McMasterAI2024-2025/TrafficLightRL.