Finite-time control based on RBF neural network for quadrotor UAVs with varied mass load

Abstract

Abstract Aiming at the problem of the gradual reduction of the weight and the external wind disturbance affect flight performance of the quadrotor Unmanned Aerial Vehicle (UAV), a dual-loop finite time control strategy based on Radial Basis Function (RBF) neural network is proposed. The UAV model under disturbance is decoupled into position outer loop subsystem and attitude inner loop subsystem. In the outer loop, the changing weight and the external wind disturbance are approximated by using RBF neural network, command filter is used to avoid the “computing explosion” problem in the traditional backstepping method, and the finite-time control method is able to improve the convergence speed of the position. In the inner loop, the cascade RBF neural network PID control which relies on the self-learning of neural network to realize the dynamic tuning of PID parameters is adopted to achieve rapid convergence of the attitude angle. The simulation results show that compared with the traditional backstepping method and cascade PID control, the convergence time is reduced by 31% on average, which verifies the superiority and effectiveness of the proposed control strategy.