Learning observer-based fault-tolerant control for semi-Markov jumping systems with probabilistic-memory event-triggered protocol.

This paper explores the problem of learning observer-based fault-tolerant control for semi-Markov jumping systems (SMJSs) with a probabilistic-memory event-triggered protocol (PMETP). The piecewise-homogeneous semi-Markov process is governed by an average-dwell-time signal, avoiding the conservatism of arbitrary switching. A novel PMETP exploits historical internal variables and accommodates random network delays, yielding fewer transmissions without sacrificing stability. A learning-based observer is introduced for non-differentiable sensor faults, enabling rapid fault error estimation. Theoretical analysis and simulation results demonstrate the effectiveness of the proposed approach in achieving fault-tolerant control for SMJSs with reduced communication burden in communication load.