CaltechAUTHORS
  A Caltech Library Service

Visuomotor Mechanical Search: Learning to Retrieve Target Objects in Clutter

Kurenkov, Andrey and Taglic, Joseph and Kulkarni, Rohun and Dominguez-Kuhne, Marcus and Garg, Animesh and Martín-Martín, Roberto and Savarese, Silvio (2020) Visuomotor Mechanical Search: Learning to Retrieve Target Objects in Clutter. In: 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE , Piscataway, NJ, pp. 8408-8414. ISBN 978-1-7281-6212-6. https://resolver.caltech.edu/CaltechAUTHORS:20210212-110435457

[img] PDF - Submitted Version
See Usage Policy.

1082Kb

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20210212-110435457

Abstract

When searching for objects in cluttered environments, it is often necessary to perform complex interactions in order to move occluding objects out of the way and fully reveal the object of interest and make it graspable. Due to the complexity of the physics involved and the lack of accurate models of the clutter, planning and controlling precise predefined interactions with accurate outcome is extremely hard, when not impossible. In problems where accurate (forward) models are lacking, Deep Reinforcement Learning (RL) has shown to be a viable solution to map observations (e.g. images) to good interactions in the form of close-loop visuomotor policies. However, Deep RL is sample inefficient and fails when applied directly to the problem of unoccluding objects based on images. In this work we present a novel Deep RL procedure that combines i) teacher-aided exploration, ii) a critic with privileged information, and iii) mid-level representations, resulting in sample efficient and effective learning for the problem of uncovering a target object occluded by a heap of unknown objects. Our experiments show that our approach trains faster and converges to more efficient uncovering solutions than baselines and ablations, and that our uncovering policies lead to an average improvement in the graspability of the target object, facilitating downstream retrieval applications.


Item Type:Book Section
Related URLs:
URLURL TypeDescription
https://doi.org/10.1109/IROS45743.2020.9341545DOIArticle
https://ieeexplore.ieee.org/document/9341545PublisherArticle
https://arxiv.org/abs/2008.06073arXivDiscussion Paper
ORCID:
AuthorORCID
Garg, Animesh0000-0003-0482-4296
Additional Information:© 2020 IEEE. We acknowledge the support of Toyota (1186781-31-UDARO). AG is supported in part by CIFAR AI chair. We thank our colleagues and collaborators who provided helpful feedback, code, and suggestions, especially Professor Ken Goldberg, Michael Danielczuk, Matt Matl, and Ashwin Balakrishna.
Funders:
Funding AgencyGrant Number
Toyota1186781-31-UDARO
Canadian Institute for Advanced Research (CIFAR)UNSPECIFIED
Record Number:CaltechAUTHORS:20210212-110435457
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20210212-110435457
Official Citation:A. Kurenkov et al., "Visuomotor Mechanical Search: Learning to Retrieve Target Objects in Clutter," 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Las Vegas, NV, USA, 2020, pp. 8408-8414, doi: 10.1109/IROS45743.2020.9341545
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:108037
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:12 Feb 2021 19:14
Last Modified:12 Feb 2021 19:14

Repository Staff Only: item control page