Leveraging Equivariant Representations of 3D Point Clouds for SO(3)-Equivariant 6-DoF Grasp Pose Generation

Byeongdo Lim, Jongmin Kim, Jihwan Kim, Yonghyeon Lee, Frank C. Park

Keywords: equivariant representations, SO(3)-equivariance, 6-DoF grasp pose generation, invariant conditional distribution on manifold, equivariant conditional manifold flows

TL;DR: An SO(3)-equivariant model for 6-DoF grasp pose generation, leveraging equivariant representations to focus on invariant conditional distribution.

Abstract: Achieving equivariance in robot learning tasks, particularly in the generation of grasp poses for various objects, has garnered significant attention due to its advantages such as data efficiency, generalization, and robustness. In this paper, we propose GraspECMF (Equivariant Conditional Manifold Flows for grasping), a novel method for SO(3)-equivariant grasp pose generation. Our method leverages SO(3)-equivariant representations of objects to learn the invariant distribution of grasp poses conditioned on the objects. Experimental validation demonstrates that our method outperforms existing methods, showcasing enhanced accuracy in grasp pose distribution learning and resulting in a higher grasp success rate.

Submission Number: 10