A Differentiable Formulation for Uncertain Pose Estimation during Contact

Jeongmin Lee, Minji Lee, Dongjun Lee

Keywords: Contact modeling, Pose estimation, Manipulation

TL;DR: ICRA 2023 Embracing contacts workshop submission

Abstract:

For many robotic manipulation and contact tasks, it is crucial to accurately estimate uncertain object poses, for which certain geometry and sensor information are fused in some optimal fashion. Previous results for this problem primarily adopt sampling-based or end-to-end learning methods, which yet often suffer from the issues of efficiency and generalizability.
In this paper, we propose a novel framework for modeling and solving this uncertain pose estimation problem in differentiable form.
To this end, we first devise a new type of geometric definition which is versatile and can provide differentiable contact features.
In conjunction with this, we develop an efficient bi-level algorithm to solve the problem.
Several scenarios are implemented to demonstrate how the proposed framework can improve existing methods.

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