Integrating Visible Light Communication and Deep Reinforcement Learning for Smarter Urban Traffic Control

Urban traffic management is increasingly challenged by rising vehicle and pedestrian flows, resulting in congestion, delays, and safety riskW. This article proposes an innovative traffic signal control framework that integrates Deep Reinforcement Learning (DRL) with Visible Light Communication (VLC) to optimize operations at intersections, which are critical bottlenecks in urban networks. A decentralized DRL agent is deployed at each intersection and trained on local traffic states, enabling real-time decision-making for both vehicular and pedestrian movements. VLC is used to support low-latency, infrastructure-to-user communication, providing accurate data on positions, speeds, queue lengths, and stop durations. The system employs a queue/request/response mechanism to adapt signal phases dynamically, resolve conflicts, and prioritize urgent flows. The proposed approach is validated through simulations and real-world trials, demonstrating superior performance over centralized and traditional agent-based methods by substantially reducing waiting and travel times while enhancing safety. The solution is scalable and adaptable to a wide range of intersection configurations, with SUMO-based experiments confirming its potential for more efficient and intelligent urban traffic control.