Fast Fractional-Order Terminal Sliding Mode Control With RBFNN Based Sliding Perturbation Observer for 7-DOF Robot Manipulator

Abstract

A new perturbation estimator, using radial basis function (RBF) neural networks (RBFNN) to modify the sliding perturbation observer (SPO), is proposed with the fast fractional-order terminal sliding mode control (FFOTSMC). It aims to control a seven-degree-of-freedom (7-DOF) robot manipulator. The new perturbation estimator applies the data-driven method RBFNN to compensate for the estimation error in the conventional SPO for the first time. The modified SPO estimates the perturbation, which contains the disturbance, dynamic uncertainties, and modeling errors. The estimated perturbation is used to design with the FFOTSMC, which improves the tracking accuracy and reduces the chattering. The FFOTSMC was designed using the fractional-order derivative to design the sliding surface and the reaching/law for reaching the sliding surface. In experiments on the robot, the proposed estimation method has been evaluated by comparing with the conventional SPO or only RBFNN with the same controller, FFOTSMC. The asymptotic stability of the controller with the new estimator is proved using Lyapunov functions for fractional-order systems.