Hard vs. Soft Contacts in Trajectory Optimization of Legged Robots: An Example with a Compliant Rubber Foot

Maximilian Raff, C. David Remy

Keywords: Soft Contacts, Hard Contacts, Legged Robots, Trajectory Optimization, Compliant Feet

TL;DR: This paper compares soft and rigid contact models in legged locomotion simulation and optimization, introducing a model of a compliant rubber foot that improves accuracy and challenges the notion that rigid models are superior.

Abstract:

In this contribution to the ICRA 2023 workshop "Embracing Contacts", we compare the use of soft and rigid contact models in the simulation and trajectory optimization of legged locomotion.
Specifically, we introduce a soft contact model that closely resembles the compliance of our legged robotic hardware [1] and carefully fit the model parameters to experimentally recorded data.
Motion simulations of the robot RAM\textit{bi} were more accurate when computed with this soft model compared to a standard rigid model.
Furthermore, we performed a simple gait optimization with both contact models for a one-legged hopper with 7 degrees of freedom, showing that the soft contact model produced results similar to those obtained with a rigid contact model.
For this particular hardware example, our work clearly breaks with the common wisdom that rigid contact models are superior in simulation and optimization of legged robotic systems.
The results presented in this contribution are taken from [2].

Confirmation: I am a student and I would like to be considered for the funding award.