一种基于正关系的多Agent协调算法

A Coordination Algorithm Between Agents Based on Positive Relations

多Agent之间的协调(coordination)与协作(cooperation)已经成为多Agent系统(multiagent system,MAS)中的一个关键问题。这是因为MAS的主要研究目标之一就是使得多Agent的信念、意图、期望、行为达到协调甚至协作。在开放、动态的MAS环境下,具有不同目标的多个Agent必须对其资源的使用以及目标的实现进行协调[1,4]。例如,在出现资源冲突时,若没有很好的协调机制,就有可能出现死锁。而在另一种情况下,当单个Agent无法独立完成目标,需要其它Agent帮助时,则需要协作。本文提出了一种基于正关系的多Agent协调机制和协调算法。在该算法中,通过使用这种协调机制,Agent能委托或接受交互中的子计划,从而形成系统负载均衡和有效降低系统运行开销。

Coordination and cooperation between agents have been a key topic in multiagent system （MAS）. Since one of the main research objectives of MAS is to obtain a coordination and cooperation between belief, intension, desire and action from different agents. In the open and dynamic MAS, agents with different goals should cooperate with each other in using resource and achieving their goals. For example, when a conflict in using resource is happening, if there lacks a good coordination mechanism, maybe a deadlock will occur in the MAS. On the other hand, when an agent can not achieve its goals independently, then it needs the help from other agents, so a cooperation mechanism is needed here. Tiffs paper proposes a coordination mechanism and a coordination algorithm between agents based on positive relations in MAS. In the algorithm, by using the coordination mechanism, an agent can consign or receive the sub-plans in its communication with other agents, so that it can enable load balancing in MAS and effectively decrease the operation cost of MAS.