The invention provides a regional traffic optimization control method and system based on multi-agent reinforcement learning, and relates to the technical field of regional traffic control, a single-agent action value network adopts centralized training and distributed execution, and the method comprises the following steps: inputting local state observation of all intersections as input into a single-agent action value network, a multi-head attention mechanism is used for distributing weights for importance degrees of all intersections at a certain T moment in a traffic area, a super network is used for fusing high-dimensional data generated by the multi-head attention mechanism, the value of all actions is output in a distributed mode, the action corresponding to the maximum value is selected, the optimal action of all the intersections under the global condition is decided, and the optimal action of all the intersections under the global condition is determined. And optimal control of regional traffic is realized.
本公开提供了基于多智能体强化学习的区域交通优化控制方法及系统,涉及区域交通控制技术领域,单智能体动作价值网络采用集中式训练、分布式执行,包括:将所有交叉口的局部状态观测作为输入,输入至单智能体的动作价值网络中,使用多头注意力机制对交通区域某T时刻的各个交叉口重要程度分配权重,利用超网络对多头注意力机制产生的高维度数据进行融合,分布式输出各个动作的价值,选取最大价值所对应的动作,决策出各个交叉口在全局下的最优动作,实现对区域交通的最优控制。
Regional traffic optimization control method and system based on multi-agent reinforcement learning
基于多智能体强化学习的区域交通优化控制方法及系统
2024-01-30
Patent
Elektronische Ressource
Chinesisch
Adaptive traffic signal control method based on multi-agent reinforcement learning
Europäisches Patentamt | 2021
|Reinforcement learning-based multi-agent system for network traffic signal control
Tema Archiv | 2010
|Traffic light signal control method based on multi-agent reinforcement learning
Europäisches Patentamt | 2022
|Multi-agent reinforcement learning method for fair adaptive traffic signal control
Europäisches Patentamt | 2023
|