Hierarchical Reinforcement Learning of Multiple Grasping Strategies with Human Instructions

Osa, T. and Peters, J. and Neumann, Gerhard (2018) Hierarchical Reinforcement Learning of Multiple Grasping Strategies with Human Instructions. Advanced Robotics, 32 (18). pp. 955-968. ISSN 0169-1864

Full content URL: https://www.tandfonline.com/toc/tadr20/current

Documents
HiPSG.pdf

Request a copy
[img] PDF
HiPSG.pdf - Whole Document
Restricted to Repository staff only until 20 August 2019.
Available under License Creative Commons Attribution-NonCommercial 4.0 International.

3MB
Item Type:Article
Item Status:Live Archive

Abstract

Grasping is an essential component for robotic manipulation and has been investi-
gated for decades. Prior work on grasping often assumes that a sufficient amount of
training data is available for learning and planning robotic grasps. However, since
constructing such an exhaustive training dataset is very challenging in practice, it
is desirable that a robotic system can autonomously learn and improves its grasp-
ing strategy. In this paper, we address this problem using reinforcement learning.
Although recent work has presented autonomous data collection through trial and
error, such methods are often limited to a single grasp type, e.g., vertical pinch
grasp. We present a hierarchical policy search approach for learning multiple grasp-
ing strategies. Our framework autonomously constructs a database of grasping mo-
tions and point clouds of objects to learn multiple grasping types autonomously. We
formulate the problem of selecting the grasp location and grasp policy as a ban-
dit problem, which can be interpreted as a variant of active learning. We applied
our reinforcement learning to grasping both rigid and deformable objects. The ex-
perimental results show that our framework autonomously learns and improves its
performance through trial and error and can grasp previously unseen objects with
a high accuracy.

Keywords:Robotic Grasping, Reinforcement Learning
Subjects:G Mathematical and Computer Sciences > G760 Machine Learning
Divisions:College of Science > School of Computer Science
ID Code:32981
Deposited On:20 Aug 2018 07:53

Repository Staff Only: item control page