Deep convolutional neural network based fractional-order terminal sliding-mode control for robotic manipulators

This paper proposes a deep convolutional neural network (DCNN) based fractional-order terminal sliding-mode (FOTSM) control strategy for tracking control of rigid robotic manipulators. The DCNN based on deep learning (DL) method is utilized to compensate the uncertainties of the system without requirement of presupposed knowledge of their upper-bounds, which makes the designed switching gain much smaller. The chattering phenomena are attenuated by combining the superiorities of a DCNN and a novel FOTSM manifold, and the control performance of rigid robotic manipulators is improved. Meanwhile, the singularity problem is solved by avoiding differentiating the exponential terms. The proposed chattering-free control strategy exhibits much robust performance against parametric uncertainties and external disturbances. Compared with the existing methods, simulations are carried out to validate the proposed methods.