Distributed State Estimation and Fault Tolerant Control of Multi-Agent System

In the realm of Distributed Multi-Agent Systems (DMAS), the topology is a crucial role as a research focal point, exerting influence upon the system's communication efficiency, collaborative capabilities and robustness. Nevertheless, relying heavily on the topology can impose limitations on the system from various perspectives. Hence, it is of paramount importance to mitigate this dependency on the topological structure within DMAS. This paper presents a novel approach based on a distributed Radial Basis Function neural network (RBFNN) observer, which utilizes the estimated states of the agents to approximate the communication information, therefore the relevant information could be acquired from neighboring agents. By utilizing Lyapunov functions, the conditions for achieving stability are derived, ensuring boundedness of errors and stability of system. This approach overcomes the reliance on the system's topology and it innovatively enables agents to acquire information from faulty agents in situations where a single agent or observer encounters a malfunction. Therefore, this method represents a new solution for distributed systems with altering or unknown global topological relations, effectively addressing node failures. Ultimately, this essay empirically validates the accuracy of communication information estimation and self-state estimation in neighbor agents when both the single agent controller and observer experience malfunctions.