Robot Action Learning Method that Corrects Errors with Gestures Published in Nikkan Kogyo Shimbun

At the 40th Annual Conference of the Japanese Society for Artificial Intelligence, Mr. Yoshimura and colleagues from our team presented their work, and the content of the presentation was published in Nikkan Kogyo Shimbun.
In robot motion learning for acquiring collaborative tasks with humans, we proposed a method that enables post hoc learning of recovery actions by additionally training on a small amount of data in which motion errors are conveyed and corrected through gestures.

For details, please refer to the article published in Nikkan Kogyo Shimbun (Japanese article).
https://www.nikkan.co.jp/articles/view/00784198

40th Annual Conference of the Japanese Society for Artificial Intelligence, Presentation Information
Yuta Yoshimura, Kento Fukasawa, Ryoichi Nakajo, Natsuki Yamanobe, Yukiyasu Domae and Tetsuya Ogata: Post-hoc Learning of Recovery Actions in Human-Robot Collaborative Tasks
https://pub.confit.atlas.jp/ja/event/jsai2026/presentation/2G6-OS-47c-05