Fault-Tolerant Model Predictive Control of a Quadrotor with an Unknown Complete Rotor Failure

In recent years, with the wide application of quadrotors in critical areas, the safety issue has attracted more and more attention and fault-tolerant control with rotor failure has become a focus of research. A fault-tolerant control (FTC) method with fault detection and diagnosis (FDD) is proposed for a quadrotor with an unknown complete rotor failure. FTC uses nonlinear model predictive control (NMPC) to achieve stable control of the position in the case of a complete rotor failure. FDD uses concurrent learning to diagnose and parameterize the unknown failure. Different from traditional linear or cascaded nonlinear fault-tolerant control methods, the proposed method does not rely on the assumption of small perturbation linearization or time-scale separation. It can autonomously implement fault-tolerant control without any failure information in priori, and can consider and deal with nonlinear dynamics and actuator constraints caused by failures. Therefore, the method has a better control effect and system bandwidth. Through simulation, the proposed method is compared with the cascaded MPC method. Under the same conditions, the cascaded MPC crashes due to the saturation of the inner loop. In contrast, the proposed method can autonomously identify the failure when an unknown complete rotor failure occurs, and achieve stable control of the position by sacrificing yaw stability.