The invention discloses a multi-agent reinforcement learning traffic signal control method based on neighbor knowledge sharing, and relates to the technical field of intelligent traffic, and the method comprises the steps: modeling a self-adaptive traffic signal control problem as a Markov decision process of neighborhood perception, modeling each intersection as an agent, and carrying out the optimization of the intelligent agent; observation and rewarding of each agent are improved based on the importance degree of local traffic information of the intersection and neighbors of the intersection; a knowledge sharing normal form between neighborhoods is established based on knowledge distillation, knowledge sharing is carried out between an intelligent agent and a neighbor network thereof by utilizing deep mutual learning, and a joint traffic signal control strategy is learned; and performing traffic signal control based on the joint traffic signal control strategy. According to the method, cooperation between intersections and neighbors connected with the intersections is considered, a knowledge sharing paradigm between neighbours is provided, deep mutual learning is carried out in the neighbours through multiple agents, the data efficiency of the algorithm is improved, and the method is very important for relieving traffic jam.
本发明公开了一种基于邻居知识共享的多智能体强化学习交通信号控制方法,涉及智能交通技术领域,包括:将自适应交通信号控制问题建模为邻域感知的马尔可夫决策过程,将每个交叉口建模为一个智能体,基于交叉口及其邻居的局部交通信息的重要程度对每个智能体的观测和奖励进行改进;基于知识蒸馏建立邻域间的知识共享范式,利用深度相互学习,在智能体及其邻居网络间进行知识共享,学习联合交通信号控制策略;基于所述联合交通信号控制策略进行交通信号控制。本发明考虑了交叉口及其连接邻居间的协作,提出了一种邻域间的知识共享范式,通过多个智能体在其邻域内进行深度相互学习,提高算法的数据效率,对缓解交通拥堵十分重要。
Multi-agent reinforcement learning traffic signal control method based on neighbor knowledge sharing
基于邻居知识共享的多智能体强化学习交通信号控制方法
2023-10-24
Patent
Electronic Resource
Chinese
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS / G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |
Adaptive traffic signal control method based on multi-agent reinforcement learning
European Patent Office | 2021
|Traffic light signal control method based on multi-agent reinforcement learning
European Patent Office | 2022
|Traffic signal control method based on multi-agent deep reinforcement learning
European Patent Office | 2025
|Multi-agent reinforcement learning traffic signal control method based on strategy multiplexing
European Patent Office | 2024
|