Tube-based robust model predictive control for fault tolerance

This paper presents a robust, fault tolerant, tube-based nonlinear model predictive controller for systems with additive external disturbances and actuator faults. The design exploits the sliding mode control design embedded in the auxiliary controller to create a lumped disturbance upper bound that represents the worst-case contribution of both faults and disturbances. In this way, the proposed design is shown to maintain robust control invariance in the presence of both forms of uncertainty. The design is expanded in two ways which utilize a double boundary layer for the sliding surface to create a blended tube design, permitting the control to take advantage of disturbance-based and lumped disturbance-based tubes. The proposed designs are implemented on the attitude control of a nanosatellite system in both simulation and experimentation, where performance is evaluated with average root mean square values on the attitude and input variables. Simulation results reveal the proposed fault-tolerant technique maintains robust control invariance in the presence of faults, unlike its nominal counterpart. Additionally, use of a double boundary layer and blended tube significantly improved tracking performance at little increase in control effort while still maintaining robust control invariance. Experimental results establish the validity of the fault-tolerant technique in practice on a model nanosatellite.