Adaptive Traffic Light Scheduling Based On Traffic Density

Traffic congestion in urban areas has intensified due to the annual growth of motorized vehicles each year. In addition, traffic congestion occurs at peak hours. This situation significantly raises fuel consumption and carbon emissions to a high level, which impacts the environment, not only in direct costs but also indirect costs such as lost time, decreased productivity, and increased driver stress. To address this issue, this research proposes adaptive traffic lights using Multi-Agent Reinforcement Learning (MARL) based on the traffic density on each road section. Each intersection is modeled as an independent agent that autonomously performs decision-making processes. To prove the effectiveness, this research presents a comparison of the average waiting time, average speed of vehicles, queue length,and throughput between fixed traffic light methods and the proposed method. Through the implementation of this method, the average waiting time of vehicle was reduce 86,67. Traffic conditions were simulated using the Simulation of Urban Mobility (SUMO) platform, which is widely employed for microscopic traffic modeling and analysis. By improving traffic signal control through MARL, this research directly support Smart Mobility pillar of the Smart City framework, which aims to leverage intelligent systems for efficient, sustainable, and adaptive urban transportation. © 2025 IEEE.