Direct adaptive fractional-order non-singular terminal sliding mode control strategy using extreme learning machine for position control of 5-DOF upper-limb exoskeleton robot systems

Abstract

This paper presents an innovative adaptive fractional-order (FO) terminal sliding mode (TSM) controller designed for a 5-degree-of-freedom (5-DOF) upper-limb exoskeleton robot. The primary focus is on addressing the challenges posed by nonlinearities and uncertainties in parameter values. The proposed approach integrates several key elements, including a novel fractional-order non-singular terminal sliding mode (FONSTSM) surface, an exponential reaching law, and the utilization of the extreme learning machine (ELM). This comprehensive strategy guarantees not only finite-time convergence but also robust stability, effectively alleviating the well-known chattering phenomenon. Furthermore, it successfully overcomes the singularity issues typically observed in conventional TSM controllers. The incorporation of the ELM with rectified linear unit (ReLU) activation function enhances robustness by facilitating the estimation of parameters related to the exponential control law. Numerical simulations provide compelling evidence of improved tracking, increased robustness against uncertainties, achievement of finite-time convergence, and notable reductions in control signal oscillations and singularity problems.