A Development of Marine Traffic Assessment Tool Constructed by Reinforcement Learning Based Search Method for Ships' Courses

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500179
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: Hiroshima City University
• Principal Investigator: KAMIO TAKESHI 広島市立大学, 情報科学研究科, 講師 (20316136)
• Co-Investigator(Renkei-kenkyūsha): MITSUBORI Kunihiko 拓殖大学, 工学部, 准教授 (10439523) ; TANAKA Takahiro 海上保安大学校, 国際海洋政策研究センター, 准教授 (20559223)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: マルチエージェントシステム / 強化学習 / 多船航路探索 / 先験的知識 / 航法 / 安全性 / 効率性

Outline of Final Research Achievements

It is very important to decide the safe and efficient courses before actual navigation. To satisfy this requirement, in this research, we have constructed the marine traffic assessment tool based on the multi-agent reinforcement learning system (MARLS) to find ships’ courses which we had proposed before. Specifically, we have proposed the extraction method for the causal information of collisions and the evaluation method for the safety and efficiency of courses. Moreover, we have modified our MARLS appropriately. Finally, it has been confirmed by numerical experiments that our MARLS can be used as the marine traffic assessment tool for the actual environment of sea area.

Free Research Field

ソフトコンピューティング，通信・信号処理