Adaptive Traffic Signal Control using Deep-Q Network for Enhanced Traffic Flow Efficiency

Traffic congestion remains a critical challenge in urban transportation, emphasizing the need for intelligent traffic control systems. Traditional control strategies, such as fixed-time and adaptive approaches, often fail to consider fluctuations in traffic flow, leading to inefficiencies as conditions worsen. This paper introduces a novel traffic signal control framework based on Deep Q-Networks (DQN), leveraging reinforcement learning to optimize signal timing dynamically. The agent interacts with its environment, incorporating real-time traffic flow, vehicle waiting times, and phase transitions to adaptively tune signals, thereby reducing congestion and improving throughput. Experiments conducted using the Simulation of Urban Mobility (SUMO) platform demonstrate significant improvements over conventional methods, reducing average waiting time to 42.5 seconds and achieving traffic efficiency of 82.2%. The scalability and flexibility of this system make it a promising solution for traffic management in densely populated cities, supporting real-time decision-making and sustainable urban mobility.