【作者】 龙晓林；

【导师】 蒋静坪；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2007， 博士

【摘要】 人工智能和机器人技术的迅猛发展给一些复杂、未知或危险性领域的任务完成带来了新的希望。自从上世纪80年末分布式移动机器人的概念被提出以后，多机器人系统的研究就如雨后春笋一般地迅速发展起来。多机器人系统在军事、航空航天、工业制造、自然界探索、灾害预防与处理、以及服务业等领域的远大应用前景吸引着大量的学者对多机器人系统的研究兴趣。但多机器人系统内部的关系极其复杂，系统的能力和潜力与系统中的各个机器人能否恰当处理好相互之间的竞争，并友好协作有着相当密切的关系。而系统中各个机器人的信息交换和共享，即各机器人间的相互通信是多机器人系统协作的充分而又必要的条件。通信的时间、效率和质量极大地影响着整个系统的适应性、灵活性和可靠性。本文在建立了多机器人系统的物理、通信、控制网络模型的基础上，利用网络图论、矩阵理论、分布式算法等一些有效的数学工具对系统的通信、信息流和分布式控制进行了分析和设计，得到了一个从通信结构、信息流传播方式到分布式控制等方面都优化的系统。本文所做的主要研究工作和成果如下：(1)建立了多机器人系统的物理、通信和控制网络模型，并分析了多机器人系统内的信息来源，概述了智能机器人的多传感器信息融合方法。(2)根据小世界网络的特点，提出了依据全局效率和局部效率演化小世界网络的方法，并按照此方法演化了不同规模的多机器人系统的通信网络，得到了一个小世界网络模型的通信结构。(3)利用网络图论、矩阵理论、分布式算法等数学工具对多机器人系统协作中的信息流和一致性问题进行了研究，多机器人系统的协作可在状态层或协作变量层达到一致性。本文给出了同步网络和部分同步网络的一致性协议，证明了所提出协议的收敛性，并通过对同步网络、部分同步网络及小世界网络的仿真进一步验证了所给出的一致性协议的收敛性。(4)研究了多机器人编队系统，建立了通信和控制网络模型，并分析了编队系统的稳定性。(5)提出了一种改进的多目标粒子群(PSO)优化算法，引入了外部知识库和变异算子，在Pareto多目标优化概念的基础上，运用此算法对多机器人编队系统进行了优化，仿真结果显示了多目标PSO算法能快速高效地优化多目标函数。对上述问题的研究结果显示，本文所建立的多机器人系统的网络模型使分析和设计多机器人系统的协作策略更简单明了；小世界网络的演化构造了一个优化的多机器人系统的通信网络；信息流的一致性收敛问题的研究证明了通信在协作中的重要性；多目标PSO算法在多机器人编队优化控制中的成功应用说明了本文所提出的改进的PSO优化算法的有效性，从而得到了一个通信、控制都优化的多机器人系统。更多还原

【Abstract】 The development of Artificial Intelligent and robotics technology brings new hopes for task implementation in complex, uncertain and hazardous environment. Since the advent of distributed mobile robotics in the late 1980s, the research field of multi-robot system (MRS) has grown dramatically, with a much wider variety of topics being addressed. The great application prospect of MRS in military, space, industry manufacture, nature exploration, disaster prevention and recovery, and personal service fields has drawn many researchers’ interest. However, because of the complexity of inter-robot relationship, the capacity and potential of MRS are associated with the conflict elimination and cooperation of inter-robot greatly. Shared information is a sufficient and necessary condition for cooperation. The efficiency and quality of communication influence the adaptation, flexibility, and reliability of multi-robot system.On the basis of physical, communication, and control network model building, the communication, information flow and distributed control methods have analyzed and designed in this thesis with some useful mathematic tools, such as algebraic graph theory, matrix theory, and distributed algorithm, and accordingly, a system was gotten with optimal communication architecture, information flow, and distributed control approaches. The main contributions of this thesis are as follows:(1) Information resource in MRS was analyzed, and multiple sensor information was fused with Support Vector Machine (SVM) method. The physical, communication and control network model in MRS were built also.(2) Under the characteristic of small-world network, a new method of small-world network evolution in terms of global and local efficiency evaluation index was proposed. Communication networks of MRS with different size were evolved, and optimal small-world network communication architecture was shaped correspondingly.(3) Information flow and consensus problems in MRS were researched with some mathematic tools, such as algebraic graph theory, matrix theory, and distributed algorithm. Consensus can be formed on the situational state or the coordination variable. In this thesis, consensus protocols of synchronous and partial synchronous networks in MRS were presented, whose convergence were proved and convergence rate were derived. The simulation results of synchronous, partial synchronous and small-world network verified further the convergence of presented consensus protocols.(4) Multi-robot formation system was studied. Under the communication and control network model, system’s stability was analyzed in terms of algebraic graph theory.(5) An improved multi-object Particle Swarm Optimization (PSO) algorithm with external repository and mutation operator was presented, and used to solve multi-object optimization problem in multi-robot formation system on the basis of Pareto optimal concept.The aforementioned research results show that the network model of MRS makes it easier to analyze and design cooperation approaches. Consensus convergence study of information flow verifies the importance of communication in multi-robot cooperation. Evolution of small-world network constitutes optimal communication architecture of MRS. Successful application of multi-object PSO algorithm in multi-robot formation optimization problem show the efficiency of improved multi-object PSO algorithm. Therefore, MRS with optimal communication and control methods is gained.更多还原