Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bioinspired composite learning control under discontinuous friction for industrial robots

Yongping Pan, Kai Guo, Tairen Sun, Mohamed Darouach

TL;DR

This work targets adaptive control of robotic systems with parameter uncertainties and discontinuous friction. It introduces a bioinspired composite error learning (CEL) framework that blends instantaneous data with data memory within a hybrid feedback-feedforward structure, enabling improved parameter estimation and tracking without reliance on high-gain control or persistent excitation. Exponential convergence of tracking and parameter estimates is established under an interval-excitation condition, and the approach includes a memory-augmented regressor, a filtered torque formulation, and a sigma-modified update to assure stability. Experimental validation on a DENSO industrial robot shows CEL achieving superior tracking accuracy and reduced control effort compared with classical FEL control, highlighting its potential for robust, energy-efficient industrial robotics.

Abstract

Adaptive control can be applied to robotic systems with parameter uncertainties, but improving its performance is usually difficult, especially under discontinuous friction. Inspired by the human motor learning control mechanism, an adaptive learning control approach is proposed for a broad class of robotic systems with discontinuous friction, where a composite error learning technique that exploits data memory is employed to enhance parameter estimation. Compared with the classical feedback error learning control, the proposed approach can achieve superior transient and steady-state tracking without high-gain feedback and persistent excitation at the cost of extra computational burden and memory usage. The performance improvement of the proposed approach has been verified by experiments based on a DENSO industrial robot.

Bioinspired composite learning control under discontinuous friction for industrial robots

TL;DR

This work targets adaptive control of robotic systems with parameter uncertainties and discontinuous friction. It introduces a bioinspired composite error learning (CEL) framework that blends instantaneous data with data memory within a hybrid feedback-feedforward structure, enabling improved parameter estimation and tracking without reliance on high-gain control or persistent excitation. Exponential convergence of tracking and parameter estimates is established under an interval-excitation condition, and the approach includes a memory-augmented regressor, a filtered torque formulation, and a sigma-modified update to assure stability. Experimental validation on a DENSO industrial robot shows CEL achieving superior tracking accuracy and reduced control effort compared with classical FEL control, highlighting its potential for robust, energy-efficient industrial robotics.

Abstract

Adaptive control can be applied to robotic systems with parameter uncertainties, but improving its performance is usually difficult, especially under discontinuous friction. Inspired by the human motor learning control mechanism, an adaptive learning control approach is proposed for a broad class of robotic systems with discontinuous friction, where a composite error learning technique that exploits data memory is employed to enhance parameter estimation. Compared with the classical feedback error learning control, the proposed approach can achieve superior transient and steady-state tracking without high-gain feedback and persistent excitation at the cost of extra computational burden and memory usage. The performance improvement of the proposed approach has been verified by experiments based on a DENSO industrial robot.
Paper Structure (8 sections, 28 equations, 4 figures, 1 table)

This paper contains 8 sections, 28 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Biological Background
  3. Problem Formulation
  4. Robot Control Design
  5. Hybrid Feedback-Feedforward Structure
  6. Composite Error Learning Technique
  7. Industrial Robot Application
  8. Conclusions

Figures (4)

  • Figure 1: A cerebellar neural circuit of simplified FEL control which is redrawn according to Wolpert1998, where the internal forward model is omitted.
  • Figure 2: A block diagram of the CEL robot control scheme.
  • Figure 3: Experimental setup: A 6-axis articulated robot. (a) A Denso robot arm (Type: VP6242G). (b) A Quanser open architecture real-time control module.
  • Figure 4: A comparison of control trajectories for the two controllers. (a) The norm of the tracking error $\mathbf e$. (b) The norm of the parameter estimate $\hat{W}$.