Derivative-free online learning of inverse dynamics models

TL;DR

This paper introduces a derivative-free online learning framework for inverse dynamics models in robotics, eliminating the need for numerical differentiation and demonstrating superior performance on real robot data.

Contribution

It proposes a novel derivative-free approach for inverse dynamics modeling, unifying various model classes and outperforming existing methods in real-world experiments.

Findings

Derivative-free methods outperform traditional approaches.

Models using the new framework achieve higher accuracy.

Experimental validation on iCub robot data confirms effectiveness.

Abstract

This paper discusses online algorithms for inverse dynamics modelling in robotics. Several model classes including rigid body dynamics (RBD) models, data-driven models and semiparametric models (which are a combination of the previous two classes) are placed in a common framework. While model classes used in the literature typically exploit joint velocities and accelerations, which need to be approximated resorting to numerical differentiation schemes, in this paper a new `derivative-free' framework is proposed that does not require this preprocessing step. An extensive experimental study with real data from the right arm of the iCub robot is presented, comparing different model classes and estimation procedures, showing that the proposed `derivative-free' methods outperform existing methodologies.