Model predictive control (MPC) has become a widely used control strategy in aerospace, process control, and automotive applications due to its optimization-based structure that allows it to account for multiple system objectives and constraints. However, uncertainties such as external disturbances and model-mismatch compromise its performance. While robust MPC (RMPC) strategies exist to counteract the effects of external disturbances, no straightforward MPC strategies exist to account for model-mismatch. A source of modelling error is actuator faults, whose impact can be mitigated by implementing fault tolerant control (FTC) strategies. Although strategies exist to imbue MPC with fault tolerance, some considerations are neglected in the current literature. First, few FTC schemes exist for explicit MPC, for which the system model is unchangeable in online implementation resulting in unavoidable model-mismatch. Second, limited schemes fundamentally change the structure of RMPC to instill them with fault tolerance. This second limitation is further complicated by the means through which uncertainty is accounted for in the control design.

This thesis further develops fault tolerant MPC literature by developing novel MPC architectures for nonlinear systems that address the aforementioned issues. The first FTC method, termed eNMPC-IMM, exploits mode detection properties of an interacting multiple model (IMM) to address the static model design of explicit nonlinear MPC (eNMPC). By equipping the eNMPC with several pre-determined fault models and weighting each by a probability variable, mode probabilities determined online by the IMM strategy permit a change in system model, reducing model-mismatch. The second FTC method, termed FT-NMPC-MPSMC, is developed in two parts. First, a tube-based RMPC method, called NMPC-MPSMC, is formed by encapsulating sliding mode control (SMC) equations within MPC resulting in a SMC that adheres to state and input constraints. To add fault tolerance, the disturbance upper bound of boundary layer SMC is restructured to incorporate an upper bound of expected fault contribution into the tube-based design. This simultaneously captures the effects of disturbances and actuator faults and maintains robust invariance of the RMPC design, which is demonstrated in simulation and in experiment. These research outcomes result in several novel contributions to the fault tolerant MPC community.