Decentralised adaptive reliable intermediate observer for state and fault estimation of distributed cyber-physical systems: a switching link-failure-allocation protocol

This paper investigates fault estimation (FE) for distributed cyber-physical systems (DCPSs) to actuator faults, sensor faults, and connection interruption faults under an undirected communication topology. A link-failure-allocation protocol is proposed to divide subsystems into different sets according to the connection types of DCPSs. To cope with dynamic changes caused by connection failures, a switching intermediate estimation algorithm with online reinforcement learning is designed to collect neighbour states and fault estimates under different link failure modes. An online reinforcement learning strategy is employed to adjust fault parameters and improve estimation accuracy. Under mixed fault conditions, a common Lyapunov function is developed to demonstrate that the global error system is uniformly and ultimately bounded (UUB). The effectiveness of the proposed method is demonstrated through simulations on an LC oscillator network.