Cooperative Control of Multi-Agent Systems

Abstract

Recent years have witnessed a lot of research efforts in the study of the coordination and control of multi-agent systems in various scientific and engineering fields (e.g. biology, computer graphics, robotics and particle physics). Some researchers are mainly interested in building mathematical models to describe the emergent collective behaviors of different biological species. Another line of research is to use multi-agent systems in addressing many industrial and military applications. These two lines of research facilitate each other in various aspects. In this dissertation, we study the cooperative control for multi-agent systems. Consensus is one of the focal points in our work. The consensus means that the state of interest of every agent in the multi-agent group reaches agreement asymptotically. This is an amazing phenomena observed in many biological groups, like bird flocks, fish schools and ant armies, and has been simulated by some famous models such as Vicsek's model [95] and Boids model [79]. For the consensus problem, first, we give relaxed sufficient conditions for the well known averaging consensus protocols. Second, we present some necessary and/or sufficient conditions for the consensus by the Vicsek's model, and then generalize our study to a class of non-averaging consensus protocols which includes Vicsek's model as a special case. Third, motivated by Vicsek's model, we provide several synchronous and asynchronous distributed algorithms that realize global, or almost global, heading consensus of multi-agent systems. Another major contribution of this thesis is on the topic of cooperative feedback control of multi-agent systems. Precisely, we propose some novel control strategies for the flocking and formation control of multi-vehicles. The main objective is to drive a team of vehicles to track some targets, force them to move along a given path or migrate to some static destination. Our design takes into account some challenges that arise in real applications, such as nonholomonic vehicle dynamics, collision avoidance and communication delay. First, we propose decentralized control laws for the leader-following flocking control of multiple point robots, and their extension for unicycle-type vehicles. Then, we design cooperative feedback controllers for the formation tracking of ground vehicle teams with collision avoidance property. Next, we put forward cooperative control strategies for the pattern preserving path following of ground vehicle teams in the presence of time-varying communication delays. The effectiveness of our proposed control methods is demonstrated in computer simulations. Numerical results are in line with the theoretical studies.