Extraction of Potentially Hazardous Situations based on Inducement of Risky Behavior: Development of Decision Support System for Assessment of the Situation

2015 Fiscal Year Final Research Report

• Project/Area Number: 25330303
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent robotics
• Research Institution: Chiba University
• Principal Investigator: ARAI Sachiyo 千葉大学, 工学(系)研究科(研究院), 教授 (10372575)
• Co-Investigator(Kenkyū-buntansha): SUZUKI Hironori 日本工業大学, 工学部, 教授 (20426258) ; MARUYAMA Yoshihisa 千葉大学, 大学院工学研究科, 准教授 (70397024)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: マルチエージェントシミュレーション / 意思決定 / 強化学習 / 認知モデル / パターン分類 / ベイジャンネットワーク

Outline of Final Research Achievements

In recent years, traffic accident of bicycle against vehicle accounts for more than 85 percent of the whole accidents. For such occasions, it is important to specify where and why accident will be happened. However, in the present circumstances, it depends on the experience without utilizing evidential data of driver’s operation.
Therefore, our study firstly analyzed logging data of human’s which were collected by using a bicycle simulator made by Honda Motor Co., Ltd., and classified them into some typical patterns. Secondly, using these patterns, we construct the Bayesian network to represent cause-and-effect of accident to implement into the multiagent simulator MATES, which can realize a mixed traffic flow, and reproduce the bicycle-caused accident scenarios. The main contribution of this study is that our constructed system can detect the unexpected accident pattern and find the causal relation via observing long period of their operation, before and after the accident.

Free Research Field

マルチエージェントシステム，強化学習，逆解析