Traffic signal lamp control method based on cooperative multi-agent reinforcement learning

The invention discloses a traffic signal lamp control method based on cooperative multi-agent reinforcement learning. The method overcomes the defect that the traditional reinforcement learning independently controls the signal lamps of each intersection and does not consider the correlation between the signal lamps, provides a double-layer coordination control strategy, and carries out coordination control on the traffic signal lamps in a road network through a local cooperation mechanism and a global cooperation mechanism. Meanwhile, according to the method, vehicle emission factors are considered, and vehicle emission can be reduced as much as possible while the road network traffic state is improved. 本发明公开了一种基于合作式多智能体强化学习的交通信号灯控制方法。该方法克服了传统强化学习单独控制各个路口的信号灯，没有考虑它们之间的相关性的缺点，提出了一种双层协调控制策略，通过局部合作机制和全局合作机制对路网中的交通信号灯进行协调控制。同时，该方法将车辆排放因素考虑在内，可以在改善路网交通状态的同时尽可能的减少车辆排放。