Dataset Information

Skill Learning by Autonomous Robotic Playing Using Active Learning and Exploratory Behavior Composition.

ABSTRACT: We consider the problem of autonomous acquisition of manipulation skills where problem-solving strategies are initially available only for a narrow range of situations. We propose to extend the range of solvable situations by autonomous play with the object. By applying previously-trained skills and behaviors, the robot learns how to prepare situations for which a successful strategy is already known. The information gathered during autonomous play is additionally used to train an environment model. This model is exploited for active learning and the generation of novel preparatory behaviors compositions. We apply our approach to a wide range of different manipulation tasks, e.g., book grasping, grasping of objects of different sizes by selecting different grasping strategies, placement on shelves, and tower disassembly. We show that the composite behavior generation mechanism enables the robot to solve previously-unsolvable tasks, e.g., tower disassembly. We use success statistics gained during real-world experiments to simulate the convergence behavior of our system. Simulation experiments show that the learning speed can be improved by around 30% by using active learning.

SUBMITTER: Hangl S

PROVIDER: S-EPMC7806109 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Skill Learning by Autonomous Robotic Playing Using Active Learning and Exploratory Behavior Composition.

Hangl Simon S Dunjko Vedran V Briegel Hans J HJ Piater Justus J

Frontiers in robotics and AI 20200403

We consider the problem of autonomous acquisition of manipulation skills where problem-solving strategies are initially available only for a narrow range of situations. We propose to extend the range of solvable situations by autonomous play with the object. By applying previously-trained skills and behaviors, the robot learns how to prepare situations for which a successful strategy is already known. The information gathered during autonomous play is additionally used to train an environment mo ...[more]

PMID: 33501210

Similar Datasets

Project description:Using robotic devices might improve recovery post-stroke, but the optimal way to apply robotic assistance has yet to be determined. The current study aimed to investigate whether training under the robotic active-assisted mode improves bimanual motor skill learning (biMSkL) more than training under the active mode in stroke patients. Twenty-six healthy individuals (HI) and 23 chronic hemiparetic stroke patients with a detectable lesion on MRI or CT scan, who demonstrated motor deficits in the upper limb, were randomly allocated to two parallel groups. The protocol included a two-day training on a new bimanual cooperative task, LIFT-THE-TRAY, under either the active or active-assisted modes (where assistance decreased in a pre-determined stepwise fashion) with the bimanual version of the REAplan® robotic device. The hypothesis was that the active-assisted mode would result in greater biMSkL than the active mode. The biMSkL was quantified by a speed-accuracy trade-off (SAT) before (T1) and immediately after (T2) training on days 1 and 2 (T3 and T4). The change in SAT after 2 days of training (T4/T1) indicated that both HI and stroke patients learned and retained the bimanual cooperative task. After 2 days of training, the active-assisted mode did not improve biMSkL more than the active mode (T4/T1) in HI nor stroke patients. Whereas HI generalized the learned bimanual skill to different execution speeds in both the active and active-assisted subgroups, the stroke patients generalized the learned skill only in the active subgroup. Taken together, the active-assisted mode, applied in a pre-determined stepwise decreasing fashion, did not improve biMSkL more than the active mode in HI and stroke subjects. Stroke subjects might benefit more from robotic assistance when applied "as-needed." This study was approved by the local ethical committee (Comité d'éthique médicale, CHU UCL Namur, Mont-Godinne, Yvoir, Belgium; Internal number: 54/2010, EudraCT number: NUB B039201317382) on July 14, 2016 and was registered with ClinicalTrials.gov (Identifier: NCT03974750) on June 5, 2019.

Dataset Information

Skill Learning by Autonomous Robotic Playing Using Active Learning and Exploratory Behavior Composition.

Publications

Skill Learning by Autonomous Robotic Playing Using Active Learning and Exploratory Behavior Composition.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets