Fault-Tolerant Aircraft Control Based on Self-Constructing Fuzzy Neural Network for Quadcopter

Abstract

This paper proposes a fault-tolerant aircraft control method based on a self-constructed fuzzy neural network for quadcopters with multiple actuator faults. We first introduce the actuator failure model and the model uncertainty. Subsequently, we establish a framework for a self-constructed fuzzy neural network observer with an adaptive rate to obtain the estimated value of the nonlinear term of the module uncertainty. We also design a multivariable sliding mode fault-tolerant controller to ensure the stability of the aircraft under this fault condition. Finally, we conduct experiments using the Pixhawk 4 flight controller installed on the QBall-X4 UAV experimental platform, such that the use of the flight controller’s fault coprocessor and redundant sensor design reduces the crash that occurs during the debugging of the control algorithm. Compared to the existing intelligent fault-tolerant control technology, our proposed method employs fewer fuzzy rules, and the number of these rules can be adaptively adjusted when the system model changes. In the experimental test, the aircraft was still able to fly stably under multi-actuator failure and interference conditions, thereby proving the stability of the proposed controller.