A diversion area mixed traffic flow control method based on graph neural network reinforcement learning comprises the following steps: step 1, setting a task scene of mixed traffic flow control, and defining a state space which can be observed by an intelligent agent; 2, constructing an adjacent matrix of a vehicle network to express a vehicle relationship, and determining an overall network structure of the model; 3, defining an action space and a reward function of the reinforcement learning agent; defining an action space as a set of lane changing actions based on a task scene, and setting a reward function to guide an intelligent agent to learn a strategy capable of obtaining an optimal action; 4, training and testing the whole model; and training the whole model by using experience playback and a target network, and then testing the performance of the reinforcement learning agent model under different traffic conditions. According to the invention, efficient control of the mixed traffic flow can be realized without occupying too many vehicle computing resources, and the traffic efficiency of the diversion area of the expressway is improved.
一种基于图神经网络强化学习的分流区混合交通流控制方法,包括:步骤1:设置混合交通流控制的任务情景,并定义智能体所能观测到的状态空间;步骤2:构建车辆网络的邻接矩阵对车辆关系进行表示,确定模型的整体网络结构;步骤3:定义强化学习智能体的动作空间和奖励函数;基于任务场景将动作空间定义为换道动作的集合,设置奖励函数以引导智能体学习能得到最优动作的策略;步骤4:对整体模型进行训练和测试;使用经验回放和目标网络对整体模型进行训练,之后在不同的交通条件下中测试强化学习智能体模型的性能。本发明能在不占用过多车辆计算资源的情况下实现对混合交通流的高效控制,提升高速公路分流区的交通效率。
Flow distribution area mixed traffic flow control method based on graph neural network reinforcement learning
基于图神经网络强化学习的分流区混合交通流控制方法
2023-05-02
Patent
Electronic Resource
Chinese
European Patent Office | 2024
|Flow distribution area traffic flow distribution method for novel mixed traffic flow
European Patent Office | 2025
|Graph neural network traffic flow prediction method based on deep learning
European Patent Office | 2023
|Traffic flow prediction method based on graph neural network
European Patent Office | 2024
|Traffic flow prediction method based on graph neural network
European Patent Office | 2025
|