A Two-Stage Learning Architecture That Generates High-Quality Grasps for a Multi-Fingered Hand

Dominik Winkelbauer   Berthold Bäuml   Matthias Humt   Nils Thuerey   Rudolph Triebel

drawing

Full Paper

Abstract

In this work, we investigate the problem of planning stable grasps for object manipulations using an 18-DOF robotic hand with four fingers. The main challenge here is the high-dimensional search space, and we address this problem using a novel two-stage learning process. In the first stage, we train an autoregressive network called the hand-pose-generator, which learns to generate a distribution of valid 6D poses of the palm for a given volumetric object representation. In the second stage, we employ a network that regresses 12D finger positions and scalar grasp qualities from given object representations and palm poses. To train our networks, we use synthetic training data generated by a novel grasp planning algorithm, which also proceeds stage-wise: first the palm pose, then the finger positions. Here, we devise a Bayesian Optimization scheme for the palm pose and a physics-based grasp pose metric to rate stable grasps. In experiments on the YCB benchmark data set, we show a grasp success rate of over 83%, as well as qualitative results on real scenarios of grasping unknown objects.

Cite this paper as:

@inproceedings{Winkelbauer2022,
    author = {Dominik Winkelbauer and Berthold B{\"a}uml and Matthias Humt and Nils Thuerey and Rudolph Triebel},
    booktitle = {IEEE International Conference on Intelligent Robots and Systems},
    title = {A Two-Stage Learning Architecture That Generates High-Quality Grasps for a Multi-Fingered Hand},
    year = {2022}}