The invention discloses a multi-agent cooperation and control method based on a traffic coordinator, provides a new benefit balancing scheme, enables agents to directly act according to team benefits, introduces a traffic coordinator network in order to realize the scheme, and enables the traffic coordinator to cooperate with the intelligent agents under the condition that the network structure is not changed. According to the method, an instruction is issued according to the importance of feature correlation of surrounding agents relative to self-agents, then the agents take actions according to the received command and the state thereof, and in addition, an anti-fact dominant function is designed to measure the influence of behaviors of the agents on team interests, so that the team interests are protected. And measuring the action ability of the agent according to the dominant function of the agent. Compared with other multi-agent algorithms, the algorithm is not only limited to the reward maximization of the agents, but measures the action rewards of the agents from the perspective of the whole group, and the expandability of the algorithm is not affected by the increase of the number of the agents.
本发明公开了一种利用基于交通协调器的多智能体合作和控制方法,本发明提出了一种新的利益权衡方案,让智能体直接按照团队利益行事,为了实现这个方案,引入了交通协调器网络,交通协调器能够在不改变网络结构的情况下,根据周围智能体相对于自我智能体的特征相关性的重要性来发布指令,然后,智能体根据收到的命令及其状态采取行动,此外,本发明设计了反事实优势函数来衡量智能体的行为对团队利益的影响,以及智能体自身的优势函数来衡量其行动能力。相较于其他多智能体算法,不仅仅只局限于智能体自身的奖励最大化,而是从整个团体的角度来衡量智能体做动作的奖励,而且算法的可扩展性不受智能体数量增加的影响。
Multi-agent cooperation and control method based on traffic coordinator
一种利用基于交通协调器的多智能体合作和控制方法
2024-06-21
Patent
Electronic Resource
Chinese
Transportation Research Record | 2005
|NTRS
|