Residual Iterative Feedback Tuning of Q-matrix for Fault-Tolerant Control against multiplicative failures

In this paper, we develop a data-driven fault tolerant control system for a discrete linear time-invariant plant with unknown multiplicative faults. The novel approach, Residual Iterative Feedback Tuning (RIFT), estimates the gradient of the design cost function by feeding back residual generation (embedded into the control parametrization) via set of experiments. Then the control performance is recovered based on an iterative descent algorithm. The main contribution of this conference paper is to present the RIFT scheme and shows its effectiveness by a simulation example.