Actuator fault tolerant control based on probabilistic ultimate bounds

In this work we introduce a novel set-based fault tolerant control scheme for linear systems under Gaussian disturbances. In the proposed strategy, actuator faults are detected and diagnosed when residual trajectories enter and remain in certain sets that are computed as probabilistic ultimate bounds. After a fault is diagnosed, the control scheme is reconfigured to take into account the corresponding actuator failure and preserve certain closed loop features. We show that our strategy can detect and diagnose the different faults considered with an arbitrarily small probability of misdetection.