Compensation of Nonlinear Torsion in Flexible Joint Robots: Comparison of Two Approaches

TL;DR

This paper compares two methods for compensating nonlinear torsion with hysteresis in flexible joint robots, improving tracking accuracy under heavy loads by using only motor-side measurements.

Contribution

It introduces and compares two novel approaches for nonlinear torsion compensation in flexible joint robots with hysteresis, using only motor-side signals.

Findings

Both methods effectively reduce torsion-induced errors.

The inverse hysteresis approach improves tracking accuracy.

Numerical example demonstrates practical applicability.

Abstract

Flexible joint robots, in particularly those which are equipped with harmonic-drive gears, can feature elasticities with hysteresis. Under heavy loads and large joint torques the hysteresis lost motion can lead to significant errors of tracking and positioning of the robotic links. In this paper, two approaches for compensating the nonlinear joint torsion with hysteresis are described and compared with each other. Both methods assume the measured signals available only on the motor side of joint transmissions. The first approach assumes a rigid-link manipulator model and transforms the desired link trajectory into that of the motor drives by using the inverse dynamics and inverse hysteresis map. The second approach relies on the modeling of motor drives and inverse hysteresis and uses the generalized momenta when predicting the joint torsion. Both methods are discussed in details along with a numerical example of two-link planar manipulator under gravity.