Backstepping control based on neural network estimation

Abstract

This paper is implementing a backstepping controller to a quadrotor that is subject to unknown disturbances and uncertainty in dynamics. Estimation and approximation of unkowns and uncertainties are performed by using Radial Base Function Neural Network (RBFNN). Along with a backstepping controller, they provide robustness to the robotic systemdespite of the presence od uncertainties. The RBFNN's output layer is utilized as an estimator then compensation eliminates the undesired effect occurs due to uncertainties. As a result, faster error convergence is achievable. A Lyapunov function was used to analyze the closed loop system, and Matlab/Simulink was used to evaluate the system performance. The demonstrated results prove the efficiency of the proposed approach.