A Nonlinear Sliding Mode Controller of Serial Robot Manipulators With Two-Level Gain-Learning Ability

Dang Xuan Ba,Joonbum Bae

doi:10.1109/access.2020.3032449

Abstract

This article presents a learning robust controller for high-quality position tracking control of robot manipulators. A basic time-delay estimator is adopted to effectively approximate the system dynamics. A low-level control layer is structured from the control error as an indirect control objective using new nonlinear sliding-mode synthetization. To realize the control objective with desired transient time, a robust sliding mode control signal is then designed based on the obtained estimation results in a high-level control layer. To promptly suppress unpredictable disturbances, adaptation ability is integrated to the controller using two-level gain-learning laws. Reaching gains and sliding gain are automatically tuned for asymptotic control performance. Effectiveness of the designed controller is concretely confirmed by the Lyapunov-based stability criterion, comparative simulations, and real-time experiments.

Highlights

Robots have been playing a key role in automation and are integral parts of developed industries such as heavy industries, mining, automobiles, construction, and consumer goods [1]-[2]
We propose a novel adaptive robust sliding mode controller for position tracking control of robot manipulators ensuring a prescribed asymptotic performance
In this paper, an adaptive robust controller is proposed for high-quality position-tracking control of robot manipulators based on a new nonlinear sliding mode scheme

Summary

Introduction

Robots have been playing a key role in automation and are integral parts of developed industries such as heavy industries, mining, automobiles, construction, and consumer goods [1]-[2]. Robots will continue to be the cornerstone of upcoming industrial revolutions, thanks to the ability of replacing humans perform production tasks and other activities with very high efficiency [3]-[5]. To accomplish such tasks, the robots need good controllers that can provide high-precision control and fast responses [3], [6]. A simpler dynamical-approximation method, naming timedelay estimation (TDE), has been recently developed to fill up such the gap [15]-[16]. Effectiveness of the TDE methods resulted in a massive number of publications [18]-[19]

Objectives

Results

Conclusion