The invention discloses a road traffic signal control optimization method based on a deep reinforcement learning algorithm. The method comprises the following steps: determining state information of an intersection according to road network environment information and intersection vehicle flow information at the current moment; a plurality of action spaces are provided on the basis of an original MUDQN algorithm, actions in different action spaces are executed under different conditions, and a new state space and a new reward function are provided. Data stored in an experience playback pool is utilized to train our model, so that a plurality of agents finally reach an NASH equilibrium state. Wherein the input parameters of the traffic signal control optimization method are intersection observation information corresponding to the signal lamp and the reward value currently obtained by the adjacent intersection, and the output parameter of the module of the traffic signal control model is the phase of the signal lamp at the current moment.
一种基于深度强化学习算法的道路交通信号控制优化方法,包括:根据路网环境信息和当前时刻路口车辆流量信息确认所述交叉口所处的状态信息;在原有MUDQN算法提出了多个动作空间,在不同的条件下,执行不同动作空间里的动作,并且提出了新的状态空间和新的奖励函数。利用经验回放池中存储的数据来训练我们的模型,使多个智能体最终达到NASH均衡状态。其中,该交通信号控制优化方法的输入参数为对应于所述信号灯的交叉口观测信息和相邻路口当前获得的奖励值,该交通信号控制模型的模块的输出参数为当前时刻所述信号灯的相位。
Multi-agent road traffic signal control method based on deep reinforcement learning algorithm
一种基于深度强化学习算法的多智能体道路交通信号控制方法
2023-10-10
Patent
Elektronische Ressource
Chinesisch
Traffic signal priority control based on shared experience multi‐agent deep reinforcement learning
DOAJ | 2023
|Europäisches Patentamt | 2024
|Traffic signal priority control based on shared experience multi‐agent deep reinforcement learning
Wiley | 2023
|