多智能体系统的聚类一致问题研究

During the past years, distributed consensus control for networks of dynamical agents has attracted many researchers from various disciplines of engineering and science due to its broad applications in various different areas such as sensor networks, rendezvous tasks, swarming and flocking models, consensus problems, and congestion control in communication networks. The existing works focus mainly on the complete consensus, meaning all the agents reach agreement about some variables of interests. This is usually not the case especially when there are more than two interacting clusters of agents. However, few results are obtained and directed towards the investigation of cluster consensus problem, even for such results, most of them are built based on simple system models and restrictive assumptions. This confines the applicability and generality of the results to capture more general and complicated scenarios in real world. This project will focus on the development of advanced control techniques in the distributed cluster consensus and coordination of interacting clusters of agents with the aim to provide solid theoretical results for investigating distributed cluster consensus behavior in multi-agent coordination. More specifically, we will seek to use tools from areas such as algebraic graph theory, matrix analysis, linear and nonlinear control theory, event-triggered data sampling method as well as the analysis and synthesis of time-delay system to address the distributed cluster consensus problems (including algorithm design, robust analysis, communication resources optimization) which arise from considering the practical engineering and complex network environments. Cluster consensus control algorithm will be designed and further systematically analyzed in view of many practical issues such as uncertainties, external disturbances, communication constraints that may happen in multi-agent systems. Finally, we aim to apply part of our theoretical results to wireless sensor networks, in particular on the multi-target tracking and clustering problems.

多智能体系统的协同一致控制作为一门热门的交叉领域在多个不同领域有着广泛的应用。目前针对多智能体系统的聚类问题还处于起步阶段，模型和研究结果大多建立个在个体系统简单动力学模型以及较理想的假设条件下，因而也限制了研究结果的一般性与适用性。本项目旨在通过对多智能体系统聚类一致问题的深入研究，从而为复杂网络化系统中出现的更具一般性的复杂群集行为-分布式聚类一致协同供理论支持。将针对由实际工程应用或复杂网络化环境下所遇到的实际问题如异类通信网络、智能体间的非线性耦合以及复杂个体动力学方程等，综合考虑各种影响系统特性的约束问题，如通讯约束、不确定性及干扰等，利用图论方法、矩阵分析、线性和非线性控制理论方法、基于事件驱动的采样系统和时滞系统分析与综合的方法解决复杂网络环境行下多智能体系统的聚类一致算法设计、网络通信资源优化、鲁棒分析等问题并将得到的理论结果运用到无线传感器网络的多目标追踪问题以及聚类问题

多智能体系统的协同一致控制作为一门热门的交叉领域在多个不同领域有着广泛的应用，但是目前相关领域的研究结果大多建立个在个体系统为简单动力学模型以及较理想的假设条件下，因而也限制了研究结果的一般性与适用性。本项目通过对多智能体系统聚类一致问题的深入研究，较为全面的探讨了异类通信网络、智能体间的非线性耦合以及复杂个体动力学方程等多种因素在多智能体系统分群演化中的作用，并综合考虑各种约束问题，如通讯约束、不确定性及干扰等，提出了一个统一的分析框架，得到了一系列物理意义明确的结论。在此基础上，进一步将多智能体系统分布式用于解决传感器网络数据聚类、时钟同步以及智能电网负荷优化等难点问题，收到了良好的效果。同时项目组构建了机器人仿真平台，尝试将分析结论用于机器人编队控制算法设计，取得了一部分成果。本项目提出了一套较为完整的多智能体系统在复杂环境下的演化分析方案，并为相关工程技术人员提供了较为实用的控制器设计方案。项目资助发表和录用包括IEEE TAC长文在内SCI期刊论文38篇，其中IEEE汇刊论文23篇(长文21篇)，Automatica论文2篇。