GenDexGrasp: Generalizable Dexterous Grasping

TL;DR

GenDexGrasp introduces a hand-agnostic, generalizable grasping algorithm trained on a large dataset, capable of rapidly generating diverse, high-success-rate grasps transferable across various robotic hands.

Contribution

It presents a novel hand-agnostic grasping method leveraging contact maps and a large multi-hand dataset, enabling rapid, diverse, and transferable grasp generation.

Findings

Achieves high success rate in grasping tasks.

Generates diverse grasping poses efficiently.

Transfers effectively among different robotic hands.

Abstract

Generating dexterous grasping has been a long-standing and challenging robotic task. Despite recent progress, existing methods primarily suffer from two issues. First, most prior arts focus on a specific type of robot hand, lacking the generalizable capability of handling unseen ones. Second, prior arts oftentimes fail to rapidly generate diverse grasps with a high success rate. To jointly tackle these challenges with a unified solution, we propose GenDexGrasp, a novel hand-agnostic grasping algorithm for generalizable grasping. GenDexGrasp is trained on our proposed large-scale multi-hand grasping dataset MultiDex synthesized with force closure optimization. By leveraging the contact map as a hand-agnostic intermediate representation, GenDexGrasp efficiently generates diverse and plausible grasping poses with a high success rate and can transfer among diverse multi-fingered robotic hands. Compared with previous methods, GenDexGrasp achieves a three-way trade-off among success rate, inference speed, and diversity. Code is available at https://github.com/tengyu-liu/GenDexGrasp.