资源受限的多Agent系统通信研究

Communicating in Resource-constrained Multi-Agent Systems

在部分可观察环境中,Agent需要知识共享完成分布式合作。在资源受限的条件下,不加选择的通信会因为消耗有用的带宽产生性能瓶颈。因此,在资源受限的系统中需要在通信效用和通信代价之间进行适当的折衷。通过对无线网络通信方法的研究,基于信息冗余的思想讨论了多Agent系统的通信问题,提出了选择性信息交换算法,保证了Agent之间必要时的通信,以此来提高合作的效率。

Agents with partial observability need to share information to achieve decentralized coordination. However,in resource-constrained systems,indiscriminate communication can create performance bottlenecks by consuming valuable bandwidth. Therefore,there is a tradeoff between the utility attained by communication and its cost. Here we addressed this tradeoff by developing a novel strategy to make communication selective based on information redundancy, ensuring communication only occurs when necessary,while maintaining acceptable coordination.