Human-Likeness Indicator for Robot Posture Control and Balance

TL;DR

This paper introduces a human-likeness indicator for assessing humanoid robot posture control and balance by comparing robot sway to human sway under perturbed conditions, aiding in robot control improvement.

Contribution

It proposes a novel performance indicator based on body sway comparison, enabling evaluation of human-like posture control in humanoid robots.

Findings

The indicator effectively differentiates robot performance from human sway.

Application demonstrates potential for guiding robot control enhancements.

The method provides a quantitative measure for human-likeness in balance control.

Abstract

Similarly to humans, humanoid robots require posture control and balance to walk and interact with the environment. In this work posture control in perturbed conditions is evaluated as a performance test for humanoid control. A specific performance indicator is proposed: the score is based on the comparison between the body sway of the tested humanoid standing on a moving surface and the sway produced by healthy subjects performing the same experiment. This approach is here oriented to the evaluation of a human-likeness. The measure is tested using a humanoid robot in order to demonstrate a typical usage of the proposed evaluation scheme and an example of how to improve robot control on the basis of such a performance indicator score