The invention discloses a traffic signal lamp timing method, device and equipment based on deep reinforcement learning, and relates to the technical field of traffic control. Green light time of each phase is pre-allocated according to the queue length of a current intersection lane, so that the total waiting time of vehicles at an intersection is effectively shortened, and the traffic signal lamp timing efficiency is improved. According to the method, before the green light time of each phase starts, the green light time of the phase is further adjusted by using a deep reinforcement learning method according to the feedback result after traffic condition interaction training, so that the optimal green light time of the current phase is predicted in real time. Furthermore, in order to better adapt to the uncertainty and complexity of the traffic flow, a phase mechanism is added in the state design, so that the situation that when the number of vehicles at different intersections is the same, the intersections are at different phase green light moments is avoided, training is more stable, and the training efficiency is improved. Therefore, the green light duration of the traffic signal lamp is reasonably distributed under the condition of complex changing traffic flow, so that the waiting time of the vehicle at the intersection is shortened.
本发明公开了基于深度强化学习的交通信号灯配时方法、装置及设备,涉及交通控制技术领域,本发明通过根据当前路口车道的队列长度预先分配各个相位的绿灯时间,以此有效降低路口车辆的总体等待时间,本发明在每个相位绿灯时间开始前用深度强化学习的方法通过与交通状况交互训练后反馈的结果好坏来进一步调整相位绿灯时长,以实现实时预测当前相位最优的绿灯时间。进一步,本发明为了能够较好地适应车流的不确定性和复杂性,在状态的设计上加入了相位的机制,以避免不同路口车辆数相同时,各路口处于不同相位绿灯时刻的情况,从而使训练更稳定,从而有效的适应在复杂变化的车流情况下合理的分配交通信号灯的绿灯时长,以降低车辆在路口的等待时间。
Traffic signal lamp timing method, device and equipment based on deep reinforcement learning
基于深度强化学习的交通信号灯配时方法、装置及设备
2023-08-15
Patent
Electronic Resource
Chinese
European Patent Office | 2024
|European Patent Office | 2022
|Traffic signal lamp control system and method based on reinforcement learning and dynamic timing
European Patent Office | 2021
|Adaptive Optimization of Traffic Signal Timing via Deep Reinforcement Learning
DOAJ | 2021
|