QDGset: A Large Scale Grasping Dataset Generated with Quality-Diversity

TL;DR

This paper introduces QDGset, a large-scale synthetic grasping dataset generated using an enhanced Quality-Diversity framework, significantly increasing the quantity and diversity of grasp data for robotic learning.

Contribution

The work extends the QD-6DoF framework with data augmentation and transfer learning, enabling scalable generation of diverse grasp datasets surpassing previous state-of-the-art.

Findings

Reduces evaluations per grasp by up to 20%

Generates 3.5 to 4.5 times more grasps and objects than prior datasets

Facilitates large-scale collaborative dataset creation

Abstract

Recent advances in AI have led to significant results in robotic learning, but skills like grasping remain partially solved. Many recent works exploit synthetic grasping datasets to learn to grasp unknown objects. However, those datasets were generated using simple grasp sampling methods using priors. Recently, Quality-Diversity (QD) algorithms have been proven to make grasp sampling significantly more efficient. In this work, we extend QDG-6DoF, a QD framework for generating object-centric grasps, to scale up the production of synthetic grasping datasets. We propose a data augmentation method that combines the transformation of object meshes with transfer learning from previous grasping repertoires. The conducted experiments show that this approach reduces the number of required evaluations per discovered robust grasp by up to 20%. We used this approach to generate QDGset, a dataset of 6DoF grasp poses that contains about 3.5 and 4.5 times more grasps and objects, respectively, than the previous state-of-the-art. Our method allows anyone to easily generate data, eventually contributing to a large-scale collaborative dataset of synthetic grasps.