The invention discloses a non-signal-lamp intersection control method and device based on deep reinforcement learning, and the method employs a deep reinforcement learning algorithm to coordinate and control the conflict-free release of an intersection, avoids the conflict release through a judgment mode of a conflict point, achieves the fairness of vehicle release through the design of a reward function, and improves the safety of a vehicle. And the release action of the target intersection without the signal lamp at the current moment is calculated through the interactive updating trained DQN network, so that the average waiting time of vehicles is shortened, the passing efficiency is improved, and the problem of vehicle congestion at the intersection without the signal lamp is solved. Compared with a non-signal-lamp intersection release method based on deep reinforcement learning in the prior art, the intersection coordination problem is formulated by using a deep learning framework, the occupation condition of a conflict area grid does not need to be recorded, and the calculation complexity is reduced; meanwhile, the coordination potential of the intersection is fully exerted by analyzing the whole intersection without signal lamps.
本发明公开了基于深度强化学习的无信号灯交叉口控制方法及设备,使用了深度强化学习算法协调控制交叉口无冲突放行,通过冲突点的判断方式避免了冲突放行,通过奖励函数的设计实现了车辆放行的公平性,以及在此基础上的最大限度放行,通过交互更新训练后的DQN网络计算得到当前时刻目标无信号灯交叉口的放行动作,减少了车辆的平均等待时间,增加了通行效率,从而解决了无信号灯交叉口车辆拥堵的问题。本发明相较于现有技术的深度强化学习的无信号灯交叉口放行方法而言,通过使用深度学习框架制定交叉口协调问题,无需记录冲突区域网格的占领情况,降低了计算的复杂度;同时,将无信号灯交叉口整体进行分析也充分发挥了交叉口的协调潜力。
Signal lamp-free intersection control method and device based on deep reinforcement learning
基于深度强化学习的无信号灯交叉口控制方法及设备
2023-08-08
Patent
Electronic Resource
Chinese
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Multi-intersection graph deep reinforcement learning signal lamp control method under data loss
European Patent Office | 2024
|European Patent Office | 2023
|Urban intersection signal control method based on deep reinforcement learning
European Patent Office | 2024
|European Patent Office | 2024
|Multi-intersection traffic signal control method based on deep reinforcement learning
European Patent Office | 2023
|