Penalty-based Numerical Representation of Rigid Body Interactions with Applications to Simulation of Robotic Grasping

TL;DR

This paper introduces a penalty-based numerical method for simulating rigid body interactions, especially in robotic grasping, addressing issues of constraint violations and improving simulation robustness.

Contribution

It presents a novel penalty-based approach that accounts for constraint violations in rigid body interactions, enhancing simulation accuracy in robotic grasping.

Findings

Effective in handling constraint violations

Improves robustness of rigid body simulations

Demonstrated through numerical experiments

Abstract

This paper presents a novel approach to numerically describe the interactions between rigid bodies, with a special focus on robotic grasping. Some of the more common approaches used to address such issues rely on satisfaction of a set of strict constraints, descriptive of the expected physical reality of such interactions in practice. However, application of constraint-based methods in a numerical setting may lead to problematic configurations in which, for instance, volumes occupied by distinct bodies may overlap. Such situations lying beyond the range of admissible configurations for constraint-based methods, their occurrence typically results in non-meaningful simulation outcomes. We propose a method which acknowledges the possibility of such occurrences while demoting their occurrence. This is pursued through the use of a penalty-based approach, and draws on notions of mechanical impedance to infer apposite reaction forces. Results of numerical simulations illustrate efficacy of the proposed approach.