The invention relates to the field of artificial intelligence and intelligent traffic, in particular to a multi-agent reinforcement learning-based vehicle-road collaborative dynamic scheduling system and method, which comprises a sensing layer, a decision-making layer and an execution layer. The sensing layer is provided with a roadside and vehicle sensing module and collects environment and vehicle information, the decision-making layer receives the information, the information is mapped to a multi-dimensional topological space through the topological dynamic state representation unit and is subjected to nonlinear dimensionality reduction to form manifold space representation, and the dual-network collaborative decision-making unit comprises a vehicle dynamic distribution network VDDPG and a roadside dynamic scheduling network RDDPG. Respectively generating a vehicle scheduling strategy and a roadside scheduling strategy, and decomposing a strategy matrix into low-rank representation; the multi-agent collaborative optimization unit constructs an agent relation graph, optimizes a communication strategy based on information entropy, deduces a coordinated scheduling strategy by using variational, and generates an optimized coordinated scheduling scheme; the execution layer receives the scheme, the vehicle execution module and the roadside control module execute instructions and feed back results to the sensing layer, and closed-loop control is formed.
多智能体强化学习的车路协同动态调度系统及其方法,本发明涉及人工智能和智能交通领域,包含感知层、决策层和执行层。感知层有路侧和车辆感知模块,采集环境与车辆信息,决策层接收这些信息,经拓扑动态状态表征单元映射到多维拓扑空间,非线性降维成流形空间表示,双网络协同决策单元含车辆动态分配网络VDDPG和路侧动态调度网络RDDPG,分别生成车辆和路侧调度策略,策略矩阵分解转为低秩表示;多智能体协同优化单元构建智能体关系图,基于信息熵优化通信策略,用变分推断协调调度策略,生成优化协同调度方案;执行层接收方案,车辆执行模块和路侧控制模块执行指令并反馈结果至感知层,形成闭环控制。
Multi-agent reinforcement learning-based vehicle-road collaborative dynamic scheduling system and method thereof
多智能体强化学习的车路协同动态调度系统及其方法
2025-05-27
Patent
Electronic Resource
Chinese
Web3.0-based multi-agent collaborative scheduling system and method
European Patent Office | 2023
|A Collaborative Control Method for Spacecraft Clusters Based on Multi Agent Reinforcement Learning
Springer Verlag | 2025
|