Journal of the Robotics Society of Japan
Online ISSN : 1884-7145
Print ISSN : 0289-1824
ISSN-L : 0289-1824
Multi-Object Grasping using Bilateral Control-Based Imitaiton Learning
Koki YamaneSho SakainoToshiaki Tsuji
Author information

2024 Volume 42 Issue 4 Pages 394-397


Imitation Learning is a method of supervised learning using human motion data to mimic human actions. It has advantages such as high sample efficiency because it learns from successful data and can obtain tasks without program modification once the data is collected. However, most robots using imitation learning operate with positional control, and it is difficult for them to perform actions that mimic the environment in tasks involving contact and passively adapt to differences in the position and shape of objects. In contrast, bilateral control-based imitation learning, which is a teleoperation technique, has been proposed as an imitation learning method that predicts force commands and performs force control. In this study, a hamburger assembly task was performed using bilateral control-based imitation learning. By evaluating the success rate of this task, we verified the effectiveness of bilateral control-based imitation learning for tasks that require handling irregularly shaped and non-rigid objects together.

Content from these authors
© 2018 The Robotics Society of Japan
Previous article Next article