Deep Reinforcement Learning-Based Traffic Signal Optimization for Smart City Transportation Systems

Abstract

Urban traffic congestion has become a significant challenge due to rapid population growth and increasing vehicle density. Conventional traffic signal control methods rely on fixed timing schedules that fail to adapt to dynamic traffic conditions. This paper proposes a deep reinforcement learning framework for adaptive traffic signal optimization in smart cities. The proposed system employs Deep Q-Networks (DQN) integrated with real-time traffic sensing technologies to continuously learn optimal signal control strategies based on vehicle density, queue length, waiting time, and traffic flow patterns. A simulation-based evaluation demonstrates that the proposed approach significantly reduces average vehicle waiting time, fuel consumption, and carbon emissions while improving overall traffic throughput. Comparative analysis against traditional fixed-time and adaptive signal control algorithms confirms the effectiveness of the reinforcement learning model. The proposed framework supports intelligent transportation systems by enabling autonomous and efficient traffic management.