Optical IoT for smart cities: integrating visible light communication and deep reinforcement learning in intelligent traffic systems

The rapid growth of urban mobility has intensified the need for intelligent and adaptive traffic management systems capable of addressing congestion, delays, and safety challenges. This study introduces an Optical Internet of Things (OIoT) framework that integrates Visible Light Communication (VLC) and Deep Reinforcement Learning (DRL) to enable decentralized, real-time optimization of traffic signals in smart cities. In the proposed architecture, each intersection operates as an autonomous DRL agent within a distributed optical IoT network. These agents process locally sensed data and cooperate through VLC-based links, acting simultaneously as sensing nodes and decision-making units. The use of VLC as a wireless optical interface ensures low-latency, highbandwidth communication between intersections, enabling real-time exchange of encoded information on vehicle queues, signal phases, and pedestrian crossings. The agents were trained using the Proximal Policy Optimization (PPO) and Double-Deep Q-Learning (DDQN) algorithms within the SUMO simulation environment, modeling modular networks of interconnected intersections adaptable to multiple urban topologies. Virtual IoT sensors captured dynamic parameters such as vehicle speed, position, queue length, and pedestrian arrival rates. The DRL agents learned to coordinate their signal phases through cooperative reward mechanisms designed to minimize waiting times, congestion, and pedestrian delays. Experimental results from 200 simulated episodes demonstrate that the proposed system outperforms centralized and noncooperative approaches, achieving reductions of up to 46.8% in average queue lengths and improvements of up to 50% in vehicle speed, while maintaining efficient pedestrian flows. VLC-based inter-agent communication enables real-time coordination between intersections, mitigating congestion and spillback without requiring centralized control. These findings highlight the potential of combining optical wireless communication with AI-driven agents to provide scalable, adaptive, and resilient solutions for next-generation smart cities, supporting sustainable and efficient urban mobility.