Machine learning-based design of a linear self-resistant attitude control system for UAV string level

Abstract This paper mainly investigates the attitude control method of the quadrotor against unknown external interference and improves the control accuracy for the subsequent design of the control algorithm by establishing a more accurate mathematical model of the quadrotor. The extended Kalman filtering algorithm is used to obtain the real-time attitude state of the vehicle for attitude solving. The inertial guidance fusion uses the Kalman filter algorithm with delay correction to estimate the vehicle’s position and velocity information. Finally, the attitude control method of serial linear self-immunity control is proposed, which estimates and compensates for the external perturbation and internal uncertainty in real-time by linear expansion observer, while the position controller is designed by using PIV control. The simulation study analyzes that this paper’s method reduces the UAV attitude angle maximum error magnitude between about 1.04° and 4.07°compared with the traditional ADRC and serial PID. The maximum tracking error of pitch angle under white noise interference is only 0.37°using the control method of this paper, and the fluctuation amplitude is reduced by 0.81 on average, which shows a stronger anti-interference ability.