Consensus Study Of Multi-agent Systems Under Independent Position And Velocity Topologies

The multi-agent system has rapidly become an crucial branch of the distributed artificial intelligence in recent twenty years. And it has also been a frontier subject at present. The multi-agent system aims at exploring large and complicated practical systems into what is small and easy to manage. Researchers mainly study such problems as the interaction between the agents, coordination and conflict resolution. And it emphasizes the intimate group cooperation among multi-agents, not the individual autonomy, fully embodying the philosophy of “the whole is greater than the sum of its parts”.So far, the multi-agent system theory has made significant achievements, especially in such a meaningful model as the second-order multi-agent system. And it has put many theoretical results into practical systems, such as unmanned aerial vehicles(UAVs) formation, coordinated control of multiple manipulators, etc.. However, most of the existing multi-agent system theory realize the consensus study in one dimensional space, and assume that the position and velocity interactions among agents are modeled by the same digraph. But in fact, such multi-agent system as coordinated control of multiple manipulators owns the nature of repetitive movements, and often requires reaching the consensus completely in the ?nite time interval, which involves the two-dimensional framework. In practical applications, the independent position and velocity topologies would, to some extent, reduce communication cost and computational complexity, optimize communication settings and reduce the rate of information interception. Therefore, we propose an iterative learning control(ILC) for the consensus of multi-agent systems in a very general framework where the position and velocity interactions among agents are modeled by independent graphs. The main contents of this dissertation are summarized as follows:1. In this paper, we propose the distributed iterative learning consensus strategy under independent position and velocity topologies and design a distributed iterative learning protocol. A sufficient condition to guarantee the multi-agent consensus is derived, leading to a reasonable cross-over study of the iterative learning control theory and the multi-agent system. Further, the proposed scheme is also extended to achieve the formation control for the multi-agent system and we get the corresponding conclusion, which is validated by experiments. Experimental results show that followers could form a desired formation under certain conditions, though not all followers can get information from the leader. It can provide an efficient and practical iterative learning control theory in some application fields like the UAVs formation, and it is of great scientific significance and application values.2. Based on the above research, a distributed iterative learning law is proposed for the second-order multi-agent system under switching independent Position and Velocity Topologies. And we get the sufficient condition to guarantee the multi-agent system consensus in the ?nite time interval, which is proved with integration by parts and contraction mapping principle. Then, the obtained conclusion is extended to the formation control, which is validated by experiments. It provides reliable theoretical basis for the multi-agent system to solve problems of topology changes caused by network communication failure.