Reliable fault-tolerant predictive control for Markov-jump Linear Systems

This work considers fault-tolerant model-predictive control for discrete-time Markov Jump Linear Systems subject to constraints on state and control variables. Hitting state constraints is a situation which should be avoided, but it is impossible to do that with total certainty when not all possible models are stable, except in trivial initial states (e.g., the origin). Hence, this work discusses how to ensure a certain degree of reliability (low likelihood of hitting constraints) in the above unstable-mode case if there exists a terminal controller which, on average, renders the system stochastically stable.