Paper Abstract and Keywords |
Presentation |
2020-03-10 16:45
Reinforcement Learning Based Multi-Ship Course Search Method with Tracking Control Hiroki Kimura, Takahiro Tomihara, Takeshi Kamio (Hiroshima City Univ.), Takahiro Tanaka (Japan Coast Guard Academy), Kunihiko Mitsubori (Takushoku Univ.), Hisato Fujisaka (Hiroshima City Univ.) NLP2019-131 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
We have developed multi-agent reinforcement learning system (MARLS) to search ships’ courses. Since the rudder angle is defined as the action in our conventional MARLS, it tends to repeat useless course search. In this report, to overcome this problem, we propose a new MARLS which has the target course as defined the action. Also, each ship traces the target course by the tracking control. Finally, it has been confirmed by numerical experiments that our proposed MARLS has better learning efficiency than our conventional MARLS. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Multi-ship course problem / Reinforcement learning / Multi-agent system / Target course / Tracking control / / / |
Reference Info. |
IEICE Tech. Rep., vol. 119, no. 471, NLP2019-131, pp. 103-108, March 2020. |
Paper # |
NLP2019-131 |
Date of Issue |
2020-03-02 (NLP) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
NLP2019-131 |
|