计算机应用研究2024，Vol.41Issue(11)：3426-3432,7.DOI:10.19734/j.issn.1001-3695.2024.03.0091
通信受限下基于注意力的多智能体协同算法
Attention-based multi-agent cooperation with limited communication
摘要
Abstract
Limited by the finite communication resources in real-world scenarios,existing multi-agent algorithms face numerous challenges in practical applications.To enhance the collaborative performance of multi-agent systems in communication-constrained environments,this paper proposed an innovative multi-agent collaborative algorithm.This algorithm was finely optimized in three aspects:communication objects,communication content,and communication time,with the aim of reducing bandwidth consumption and improving system collaboration performance.Firstly,the algorithm introduced a scalable multi-dimensional information encoding and fusion module,which possessed excellent information extraction and fusion capabilities.By encoding multi-dimensional information,this module effectively extracted information such as agent perception and inter-agent communication,and achieved information complementarity during the fusion process,thereby enhancing the accuracy and efficiency of collaborative decision-making.Secondly,the algorithm included a gate-controlled communication decision module,which possessed adaptive information communication and decision-making capabilities,allowing it to decide whether to communicate based on the value of communication information.Experimental results demonstrate that compared to contrastive algorithms,the proposed algorithm exhibits superior collaborative performance in scenarios such as collaborative navigation and collaborative inspection.Furthermore,the algorithm significantly reduces the amount of communication information among agents,further improving its adaptability in complex environments.关键词
多智能体系统/多智能体协同/多智能体强化学习/通信资源受限/协同算法Key words
multi-agent system/multi-agent cooperation/multi-agent reinforcement learning/limited communication resource/cooperation algorithm分类
信息技术与安全科学引用本文复制引用
杨桂松,魏满舟..通信受限下基于注意力的多智能体协同算法[J].计算机应用研究,2024,41(11):3426-3432,7.基金项目
国家自然科学基金资助项目(61602305,61802257) （61602305,61802257）
上海市自然科学基金资助项目(18ZR1426000,19ZR1477600) （18ZR1426000,19ZR1477600）
