The invention discloses an intersection reinforcement learning signal control method based on network connection vehicle detection state perception, and the method comprises the steps: enabling an intersection entrance lane to be divided into a certain number of grids under the condition that a network connection automatic driving vehicle has a certain market permeability, taking the network connection automatic driving vehicle as a mobile sensor, and carrying out the recognition of the network connection automatic driving vehicle. Detecting data of real-time positions, speeds and the like of surrounding vehicles, and filling the data into corresponding entrance lane grids to form a grid filling matrix; a signal lamp is regarded as an intelligent agent, a Markov decision process is designed, a grid filling matrix is regarded as an intelligent agent state, phase switching is regarded as action, and vehicle updating waiting time is regarded as a reward function. A depth strategy gradient algorithm is adopted to train an intelligent agent, the waiting time of vehicles can be shortened, and optimization of the queuing length and the average speed is achieved at the same time. According to the method provided by the invention, traffic congestion can be reduced when the signal traffic port is controlled, and traffic delay, carbon emission and energy consumption caused by the traffic congestion are relieved.
本发明公开了一种网联车探测状态感知的交叉口强化学习信号控制方法,在网联自动驾驶车辆具有一定市场渗透率的情况下,将交叉口进口道划分为一定数目的网格,以网联自动驾驶车辆作为移动传感器,探测周围车辆的实时位置、速度等数据,将数据填充到相应的进口道网格中,形成网格填充矩阵。将信号灯视作智能体并设计马尔可夫决策过程,以网格填充矩阵作为智能体状态,以相位切换为动作,以车辆更新等待时间为奖励函数。采用深度策略梯度算法训练智能体,可以减少车辆的等待时间,同时实现排队长度以及平均速度的优化。本发明提出的方法能够在对信号交通口进行控制时减少交通拥堵,缓解因交通拥堵带来的交通延误和碳排放及能源消耗。
Intersection reinforcement learning signal control method for network connection vehicle detection state perception
一种网联车探测状态感知的交叉口强化学习信号控制方法
2022-04-15
Patent
Electronic Resource
Chinese
Intersection intelligent network connection vehicle control method based on reinforcement learning
European Patent Office | 2024
|European Patent Office | 2023
|European Patent Office | 2024
|Urban intersection signal control method based on deep reinforcement learning
European Patent Office | 2024
|Multi-Intersection Signal Control Based on Asynchronous Reinforcement Learning
DOAJ | 2025
|