Optimal Gain Scheduling for Fault-Tolerant Control of Quadrotor UAV Using Genetic Algorithm-Based Neural Network

Quadrotors have been widely utilized in commercial and industrial applications as unmanned aerial vehicles thanks to their mobility, simple design and accessibility. However, quadrotors are prone to actuator faults that hinder their flight capability and can ultimately render the flight control system ineffective. Fault-tolerant control techniques have been developed to compensate for the impacts of faults in quadrotor’s performance with fault-tolerant PID control being a frequent control design in this field. Despite the advantages, PID control suffers limitations in quadrotor systems given their high nonlinearity and underactuation in addition to the controller’s poor performance for broad fault scenarios without active tuning in its control gains. In this paper, a gain-scheduled fault-tolerant PID control method is proposed which uses a combination of genetic algorithm (GA) and artificial neural network (ANN) techniques for optimal PID tuning to address quadrotor flight performance with one or multiple damaged rotors. In the proposed design, the GA is applied to optimally tune PID control parameters based on different fault scenarios, and the acquired set of tunings are used to train an ANN that will be as an online inference model for optimal PID tuning and the detected faults and tracking errors. Simulation results are presented showing the performance of the proposed fault-tolerant PID controller. The proposed design is compared to a standard fuzzy gain-scheduled fault-tolerant PID through simulation analysis showing the superior performance of the proposed work in both quadrotor position and attitude controls and handling loss of effectiveness actuator faults.