[공지사항]을 빙자한 안부와 근황 
Show more

Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Fast Bilateral Teleoperation and Imitation Learning Using Sensorless Force Control via Accurate Dynamics Model

Created by
  • Haebom

Author

Koki Yamane, Yunhan Li, Masashi Konosu, Koki Inami, Junji Oaki, Sho Sakaino, Toshiaki Tsuji

Outline

This paper presents a four-channel bidirectional control system for high-speed remote manipulation using a low-cost manipulator. While conventional one-way control methods have difficulty in high-speed or contact-intensive tasks due to the lack of force feedback, the system in this paper provides force feedback through nonlinear term compensation, velocity and external force estimation, and variable gain control according to inertia change based on the manipulator dynamics model without force sensors. In addition, it shows performance improvement by integrating force information into the input and output of reinforcement learning policy using collected data. Through this, we emphasize the practicality of high-precision remote manipulation and data collection with low-cost hardware.

Takeaways, Limitations

•
Takeaways:
◦
Presenting the possibility of high-speed remote manipulation using low-cost force sensorless manipulators.
◦
Effectively implement force feedback through 4-channel bidirectional control.
◦
Achieving performance improvements by incorporating force information into reinforcement learning.
◦
Demonstrating the potential for high-precision remote operation and data collection with inexpensive hardware.
•
Limitations:
◦
The generalization performance of the proposed method needs to be verified for various tasks and environments.
◦
A clear description of the characteristics and limitations of the manipulator used is needed.
◦
Applicability and performance evaluation for more complex tasks are needed.
◦
Additional research is needed for application to real industrial environments.
👍
Made with Slashpage