Learning-based Motion Stabilizer Leveraging Offline Temporal Optimization

Min Sung Ahn, Hosik Chae, Colin Togashi, Dennis Hong, Joohyung Kim, Sungjoon Choi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

During loco-manipulation, instabilities to the robot's base can be introduced by the manipulator's motions. Trajectories that are generated on-the-fly may jeopardize the stability and safety of the robot and its surroundings. This work proposes a self-supervised learning-based pipeline to keep a robot stable while executing a given trajectory. Empirical results show that the desired objective can be achieved with the proposed pipeline. Experiments are done in simulation and on hardware on a unique multi-modal, manipulation-capable legged robot, and its scalability is tested on a conventional manipulator.

Original languageEnglish (US)
Title of host publication2022 19th International Conference on Ubiquitous Robots, UR 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages129-136
Number of pages8
ISBN (Electronic)9781665482530
DOIs
StatePublished - 2022
Event19th International Conference on Ubiquitous Robots, UR 2022 - Jeju, Korea, Republic of
Duration: Jul 4 2022Jul 6 2022

Publication series

Name2022 19th International Conference on Ubiquitous Robots, UR 2022

Conference

Conference19th International Conference on Ubiquitous Robots, UR 2022
Country/TerritoryKorea, Republic of
CityJeju
Period7/4/227/6/22

ASJC Scopus subject areas

  • Artificial Intelligence
  • Hardware and Architecture
  • Mechanical Engineering
  • Control and Optimization

Fingerprint

Dive into the research topics of 'Learning-based Motion Stabilizer Leveraging Offline Temporal Optimization'. Together they form a unique fingerprint.

Cite this