【作者】 贺晨龙；

【导师】 张继业；

【作者基本信息】 西南交通大学 ， 载运工具运用工程， 2009， 硕士

【摘要】 在自然界中,人们经常可以观察到一种迷人的现象,大量的鸟、鱼、昆虫等生物会形成一个规则有序的群体进行运动。由大量这样个体抽象而成的系统被称为多智能体系统(Multi-Agent Systems,MAS),多智能体系统的协调问题成为了当前的一个研究热点。本文针对多智能体系统协调运动的同步条件、多智能体系统群集控制和多机器人系统编队控制问题进行了研究。首先,对具有位置依赖约束的Vicsek模型进行了仿真分析,观察到了智能体之间的信息交换拓扑图并不是系统同步的唯一约束条件,为后续的理论证明打下了基础。其次,针对具有一阶积分器模型的多智能体系统,利用神经网络模型和复杂网络牵引控制方法设计了基本群集控制器、满足作动器饱和约束的群集控制器和时滞条件下的群集控制器,并通过理论证明和数值仿真对控制器进行了验证。最后,针对具有二阶积分器模型的多机器人系统,利用人工势能场和虚拟领导者方法设计了两种编队控制器,第一种控制器能够满足做动器饱和约束,第二种控制器能够同时保证机器人达到期望间距和方位角,并通过理论证明和数值仿真验证了控制器的有效性。通过本文的研究工作,为多智能体系统同步条件的理论证明提供了数值依据。首次将神经网络稳定性和复杂网络牵引控制和多智能体系统群集控制问题相结合,改进了前人提出的多机器人编队的人工智能场函数,具有一定的理论和工程价值。更多还原

【Abstract】 In nature, a fascinating phenomenon has usually been observed that a flock of birds, a school of fish and a colony of insects etc. form a regular, ordered collection to move. These systems consist of plenty of individuals is called Multi-Agent System, MAS for short. The coordination of MAS has become a hot topic in research recently.In this paper, we focus on the problems of the coordination condition of Multi-Agent Systems, the flocking control of Multi-Agent Systems and the formation control of Multi-Robot Systems. First, we simulate the Vicsek model with the constraint of position dependence. The results show that the connected condition of information topology is not the only constraint of model synchronization, which is the numerical foundation of future work. Second, for Multi-Agent Systems with first integrator model, using the model of neural networks and pinning control of complex networks, we design the basic flocking controller, flocking controller with constraint of actuator saturation and flocking controller with time-delayed information topology. Moreover, through theoretical proof and numerical simulation, the effectiveness of controllers is validated. Finally, for Multi-Robot Systems with double integrator model, using artificial potential field and virtual leader theory, we present two kinds of controller, one of which can meet the constraint of actuator saturation, the other of which can guarantee the desired distance and angle between two robots at the same time, and verify the controllers with theories and simulations.The research in this paper gives the numerical proof of convergence of Multi-Agent Systems. We relate the stability of neural network and pinning control of complex network to the flocking control of Multi-Agent Systems for the first time, and also improve the previous functions of artificial potential field for the formation of Multi-Robot Systems. All the work in paper has some value in theory and engineering.更多还原