The invention discloses a reinforcement learning intelligent traffic signal lamp control method based on a double-strategy network, and the method comprises the steps: 1, defining a satisfaction index, depicting the vehicle driving quality at a vehicle individual level, and carrying out the accurate modeling of an intersection traffic condition at a global level; 2, selecting a proper intersection signal lamp phase and corresponding duration according to the modeling of the intersection traffic condition in the step 1 by using a variable-duration traffic signal control method of a double-strategy network; and step 3, designing a state and an award in the reinforcement learning method based on the satisfaction index in the step 1, designing an action in the reinforcement learning method based on the double-strategy network in the step 2, using a reinforcement learning agent of each intersection to use a Deep Q Network reinforcement learning algorithm with two strategy networks, and performing real-time control on traffic lights according to the traffic flow condition of the intersection. The reinforcement learning agent can rapidly converge to a good control strategy, and the learning speed and the control quality of the method are better than those of an existing method.
本发明公开了一种基于双策略网络的强化学习智能交通信号灯控制方法,包括:步骤1:定义满意度指标,在车辆个体层面刻画车辆行驶质量,在全局层面对路口交通情况进行准确建模;步骤2:利用双策略网络的可变时长交通信号控制方法根据步骤1中对路口交通情况的建模选择合适的路口信号灯相位与相应的持续时间;步骤3:基于步骤1的满意度指标设计强化学习方法中的状态与奖励,基于步骤2中的双策略网络设计强化学习方法中的动作,利用每个路口的强化学习智能体使用带有两个策略网络的Deep Q Network强化学习算法,根据路口车流情况对交通信号灯实时控制。本发明的强化学习智能体可以快速收敛到一个好的控制策略,在学习速度与控制质量上均优于现有方法。
Reinforced learning intelligent traffic signal lamp control method based on double-strategy network
基于双策略网络的强化学习智能交通信号灯控制方法
2022-11-01
Patent
Elektronische Ressource
Chinesisch
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Traffic signal lamp intelligent control method, storage medium and traffic signal lamp
Europäisches Patentamt | 2020
|Europäisches Patentamt | 2023
|