|
JAIST Repository >
b. 情報科学研究科・情報科学系 >
b11. 会議発表論文・発表資料等 >
b11-1. 会議発表論文・発表資料 >
このアイテムの引用には次の識別子を使用してください:
http://hdl.handle.net/10119/19674
|
| タイトル: | Explainable Deep Reinforcement Learning for Patrol Speed Control of Rail-Guided Robot System |
| 著者: | Lee, Hosun
Kwon, Jaesung
Chong, Nak Young
Yang, Woosung |
| キーワード: | Rail-guided robot
Safety patrol robot
Explainable AI
Reinforcement learning
Deep Deterministic Policy Gradient algorithm
Patrol speed control |
| 発行日: | 2024-12-04 |
| 誌名: | The 12th International Conference on Robot Intelligence Technology and Applications (RiTA 2024) |
| 抄録: | Intelligent facility management systems can operate autonomously, reducing the workload for workers. However, existing machine learning-based systems do not provide users with information that serves as an understandable basis for decisions on tasks performed autonomously, making it difficult for users to trust the system. Facility management systems, which require high reliability in safety management, need functions that explain autonomously performed tasks as autonomy increases to support user understanding. This paper proposes an explainable deep reinforcement learning framework for a rail-guided patrol robot, utilizing the Deep Deterministic Policy Gradient (DDPG) algorithm and the Grad-CAM algorithm. The DDPG-based framework autonomously controls the patrol speed of the robot based on environmental information at the site. The Grad-CAM-based XAI technique is applied to the patrol speed control framework to visually highlight environmental factors that significantly impact the robot’s decision-making. Through the proposed safety patrol robot control framework, the robot adjusts its speed in response to changes in the patrol environment, projecting the specific locations that influenced speed changes onto image data at the points where these changes occurred. |
| Rights: | This is the author's version of the work. Copyright (c) Author(s). The 12th International Conference on Robot Intelligence Technology and Applications (RiTA 2024). Personal use of this material is permitted. This material is posted here with permission of RiTA 2024. |
| URI: | http://hdl.handle.net/10119/19674 |
| 資料タイプ: | author |
| 出現コレクション: | b11-1. 会議発表論文・発表資料 (Conference Papers)
|
このアイテムのファイル:
| ファイル |
記述 |
サイズ | 形式 |
|---|
| N-CHONG-I-0130.pdf | | 3280Kb | Adobe PDF | 見る/開く |
|
当システムに保管されているアイテムはすべて著作権により保護されています。
|
