摘要
分布式多智能体系统(MAS)一致性是实现协调控制的首要条件。多智能体一致性协议通常使用一跳邻域信息进行收敛,每个智能体仅通过有限的局部信息导致收敛速度缓慢。为解决上述问题提出了一种基于二跳邻居的分布式一致性协议。首先,提出方法综合一跳和二跳的邻域信息作为计算智能体的下一状态的基础,增强了智能体的决策能力;其次,提出了一种关键邻居选择方法,解决了收集二跳邻域信息可能导致的对计算效率和鲁棒性的影响;然后,设计了基于相对最近邻拓扑的约束集构造方法以增大约束集的范围。最后,通过引入凸包和实例分析,证明了该协议的连通性和收敛性,同时对多种类型的拓扑进行了大量的仿真实验。实验结果表明,该协议在各种拓扑下都具有优势,特别是在大规模高密度拓扑上。
Consensus in distributed MAS serves as the fundamental prerequisite for achieving coordinated control.Tradi-tional multi-agent consensus protocols typically utilize one-hop neighborhood information for convergence,where each agent relies solely on limited local information,resulting in slow convergence rates.To address these challenges,this paper proposed a distributed consensus protocol based on two-hop neighbor information.Firstly,the proposed method integrated the one-hop and two-hop neighborhood information as the basis for computing agents’next states,thereby enhancing agents’decision-making capabilities.Secondly,it developed a key neighbor selection method to mitigate potential impacts on computational efficiency and robustness caused by collecting two-hop neighborhood information.Furthermore,it designed a constraint set construction method based on relative nearest-neighbor topology to expand the scope of constraint sets.Finally,through the introduction of convex hull theory and case analysis,this paper rigorously proved the protocol’s connectivity and convergence.Extensive simulation experiments across various topology types demonstrate that the proposed protocol exhibits superior performance under diverse topological configurations,particularly showing significant advantages in large-scale high-density topologies.
作者
谢光强
冯衍达
李杨
Xie Guangqiang;Feng Yanda;Li Yang(School of Computer Science&Technology,Guangdong University of Technology,Guangzhou 510006,China)
出处
《计算机应用研究》
北大核心
2025年第8期2482-2489,共8页
Application Research of Computers
基金
国家自然科学基金资助项目(62006047)。
关键词
多智能体系统
二跳邻域
一致性
大规模混合拓扑
分布式控制
multi-agent system(MAS)
two-hop neighborhood
consensus
largescale hybrid topology
distributed control
