Development of a Robust-Adaptive Fault-Tolerant Control Algorithm Enhanced by Data-Driven Actuator Fault Estimation for a Multirotor UAV

Existing fault-tolerant control methods for multicopter UAVs often exhibit degraded performance under actuator faults and modeling uncertainties; therefore, this paper presents a robust-adaptive control algorithm for multicopter UAVs operating under actuator fault conditions. A data-driven approach based on an Artificial Neural Network (ANN) is employed to estimate actuator fault using IMU measurements. The ANN is trained using data generated from a closed-loop system controlled by a robust-adaptive control, rather than an open-loop configuration, improving its ability to capture realistic fault dynamics. To mitigate limitations in training data coverage, an adaptive mechanism is incorporated to enhance robustness under varying operating conditions. In addition of inherent fault-tolerant control characteristics of the robust-adaptive control, the estimated fault signals are used for motor speed compensation to enhance the robustness of the algorithm. The inner-loop controller is designed based on a robust-adaptive algorithm, ensuring system stability and robustness against model uncertainties and fault estimation errors, even the actuator faults change both system dynamics and actuation. The outer-loop Proportional–Integral–Derivative (PID) controller is employed to achieve accurate trajectory tracking. For validation and benchmarking, a standalone robust-adaptive controller and a model-based recursive least squares (RLS) estimator are also implemented. Simulation results demonstrate that the proposed ANN-based approach provides accurate fault estimation and effective compensation, resulting in improved tracking performance under actuator fault conditions. Furthermore, the proposed framework contributes to the development of a fault-tolerant UAV systems by integrating robust-adaptive control, ANN-based fault estimation, and actuator compensation into a unified architecture, thereby enhancing reliability, robustness, and tracking performance in the presence of actuator faults and modeling uncertainties.