Fault-Tolerant Model Predictive Control of a Fixed-Wing UAV with Actuator Fault Estimation

Abstract

The vast majority of today’s engineering systems possess operational constraints and have multiple inputs and outputs. This classifies them as Multi-Input Multi-Output (MIMO) systems. This paper develops a novel observer-based fault diagnosis scheme with the capability of simultaneous state and actuator fault estimation for Linear Time-Invariant (LTI) MIMO systems, which is then integrated with Model Predictive Control (MPC) method for achieving fault-tolerant control. The application within this study is chosen to be the longitudinal flight control of a fixed-wing Unmanned Aerial Vehicle (UAV). The observer-based method is combined with two MPC schemes to detect and compensate randomly occurring actuator faults in real time. The faults are modeled as a Loss Of Effectiveness (LOE). For the first (efficient) MPC method, a simple reconfiguration can be performed in the event of faults, as it is based on an absolute input formulation. However, as the second (integral-action) MPC is based on an incremental input formulation, reconfiguration is not required, since this algorithm has a degree of implicit fault tolerance. Numerical simulations demonstrate the effectiveness of the proposed approach for both MPC schemes.