Adaptive and Responsive Traffic Signal Control Using Reinforcement Learning and Fog Computing

As urban traffic becomes increasingly complex, conventional actuated traffic signal control methods are showing their limitations in mitigating congestion, and Deep Reinforcement Learning (DRL) is considered a promising solution. Existing studies mainly focus on using distributed agents, each intersection being controlled by a separate agent that can communicate with its neighbors to facilitate local coordination, which subsequently increases network complexity and usage. This article presents a DRL-based intelligent traffic signal control framework that leverages the fog computing paradigm. Traffic signals are divided into clusters, each controlled by a shared DRL agent in a fog node based on the collective information of the regional traffic situation. We have conducted simulation experiments in multiple scenarios, and the experimental results show that the proposed method yields lower average waiting times compared to existing methods.