Traffic Signal Control via Reinforcement Learning: A Review on Applications and Innovations

Traffic signal control plays a pivotal role in intelligent transportation systems, directly affecting urban mobility, congestion mitigation, and environmental sustainability. As traffic networks become more dynamic and complex, traditional strategies such as fixed-time and actuated control increasingly fall short in addressing real-time variability. In response, adaptive signal control—powered predominantly by reinforcement learning—has emerged as a promising data-driven solution for optimizing signal operations in evolving traffic environments. The current review presents a comprehensive analysis of high-impact reinforcement-learning-based traffic signal control methods, evaluating their contributions across numerous key dimensions: methodology type, multi-agent architectures, reward design, performance evaluation, baseline comparison, network scale, practical applicability, and simulation platforms. Through a systematic examination of the most influential studies, the review identifies dominant trends, unresolved challenges, and strategic directions for future research. The findings underscore the transformative potential of RL in enabling intelligent, responsive, and sustainable traffic management systems, marking a significant shift toward next-generation urban mobility solutions.