Bimanual Grasp Synthesis for Dexterous Robot Hands

TL;DR

This paper introduces BimanGrasp, a novel algorithm for synthesizing bimanual grasp poses for dexterous robot hands, supported by a large dataset and a diffusion model that improves speed and success rate.

Contribution

It presents the first large-scale bimanual grasp dataset and a diffusion-based model for faster, more successful grasp synthesis in robotic manipulation.

Findings

Achieved a 69.87% success rate with BimanGrasp-DDPM.

Created over 150,000 verified grasps on 900 objects.

Significantly accelerated grasp synthesis compared to previous methods.

Abstract

Humans naturally perform bimanual skills to handle large and heavy objects. To enhance robots' object manipulation capabilities, generating effective bimanual grasp poses is essential. Nevertheless, bimanual grasp synthesis for dexterous hand manipulators remains underexplored. To bridge this gap, we propose the BimanGrasp algorithm for synthesizing bimanual grasps on 3D objects. The BimanGrasp algorithm generates grasp poses by optimizing an energy function that considers grasp stability and feasibility. Furthermore, the synthesized grasps are verified using the Isaac Gym physics simulation engine. These verified grasp poses form the BimanGrasp-Dataset, the first large-scale synthesized bimanual dexterous hand grasp pose dataset to our knowledge. The dataset comprises over 150k verified grasps on 900 objects, facilitating the synthesis of bimanual grasps through a data-driven approach. Last, we propose BimanGrasp-DDPM, a diffusion model trained on the BimanGrasp-Dataset. This model achieved a grasp synthesis success rate of 69.87\% and significant acceleration in computational speed compared to BimanGrasp algorithm.