关键词: 状态-动作-回报-状态-动作, 实时交通信号控制, 强化学习, 交通评价指标, 时序差分学习

Abstract: In view of the existing traffic light controller lack of the past experience of learning ability, unable to adapt to the dynamic change of the actual traffic environment, this paper proposes a SARSA（λ）-based control model for real-time traffic signal, and gives a traffic signal optimization model and algorithm. The model uses reinforcement learning algorithm to obtain the optimized scheduling strategy for traffic control. Simulation results show that the proposed model is superior to the existing traffic control model, and can improve the real-time dynamic traffic control to achieve better.

Key words: State-Action-Reward-State-Action（SARSA）, real-time traffic signal control, reinforcement learning, traffic evaluation indicators, temporal difference learning