Consensus Of Multi-Agent Systems With Actuator Faults

Multi-agent system is a kind of complex network systems, which is composed of multiple individuals through the communication topology. Agents in the system can reach the overall goals by the local communication and distributed control protocols,such as the fish and the birds besiege the predators to avoid attack, etc. Many practical systems can be modelled by the multi-agent system, therefore the multi-agent system has the important research significance. Consensus is the most basic aspect of multiagent system. This thesis mainly studies the issue of fault tolerant consensus, namely,the consensus under actuator faults.First, the fault tolerant consensus of linear multi-agent systems with the undirected communication topology is considered. Against actuator failures, we establish a general fault model, which covers all the possible fault modes- partial actuator faults and the outage case. Based on this model, two fault tolerant control protocols, a fixed-gain control protocol and an adaptive-gain control protocol are designed. Moreover, we derive sufficient conditions which guarantees the fault tolerant consensus.Second, for multi-agent systems with a leader, we deal with the fault tolerant consensus tracking problem of the linear dynamics and the dynamics including Lipschitz nonlinear parts. By the designed adaptive-gain control protocol, the consensus tracking problem in the presence of actuator faults is solved for the case where the leader has a bounded control input leading the followers' states access to the expected trajectories.Finally, the observer is utilized to analyze the fault tolerant consensus problem of linear multi-agent systems subject to actuator faults. Some sufficient conditions are proposed by the Lyapunov method to achieve consensus.