A Survey on Reinforcement Learning Techniques for Urban Traffic Signal Control

With the increasingly serious problem of urban traffic congestion, adaptive Traffic Signal Control (TSC) systems based on Reinforcement Learning (RL) have gradually become a research hotspot for solving the complex dynamics of modern traffic flow. Traditional TSC methods show obvious limitations in dealing with rapidly changing traffic flows, such as response hysteresis and poor local coordination. In this paper, the application of RL in urban TSC is discussed in depth, covering the evolution of core theories and key algorithms, and the key techniques of RL in TSC modeling are analyzed in detail, such as the state representation, action space design, and the method of constructing reward functions. Meanwhile, the article also explores the potential of Multi-Agent Reinforcement Learning (MARL) in area cooperative control, and looks forward to the future fusion of cutting-edge technologies such as large language modeling. This paper aims to provide researchers in the field of TSC with a comprehensive technical reference.