A Proposal for an Autonomous Navigation Learning Method via Neural Radiance Fields that Does Not Require Actual Robot
-
- AOKI Junki
- Ricoh Co., Ltd. Kyushu University
-
- SASAKI Fumihiro
- Ricoh Co., Ltd.
-
- MATSUMOTO Kohei
- Kyushu University
-
- YAMASHINA Ryota
- Ricoh Co., Ltd.
-
- KURAZUME Ryo
- Kyushu University
Bibliographic Information
- Other Title
-
- Neural Radiance Fieldsを用いた実機不要な自律移動学習手法の提案
Description
<p>This paper investigates using Neural Radiance Fields (NeRF) to enable autonomous navigation simulations without the need for actual robots. NeRF's strength lies in its ability to render photorealistic images, promising a solution to the long-standing challenge of the domain gap between simulation and real-world environments. We present findings that validate the effectiveness of a NeRF-simulated environment for training a reinforcement learning policy. Once trained in the NeRF environment, this policy can navigate an actual robot in the real world.</p>
Journal
-
- The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec)
-
The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) 2024 (0), 1P1-M09-, 2024
The Japan Society of Mechanical Engineers
- Tweet
Keywords
Details 詳細情報について
-
- CRID
- 1390021149548883328
-
- ISSN
- 24243124
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
- Crossref
-
- Abstract License Flag
- Disallowed