ファジィモデル自動車の再構成データを与える実測データ規範シミュレーションの提案と検討

概要

論文の詳細を見る
We have proposed a microscopic road traffic simulator (MITRAM) for analyzing chronic traffic congestion on road in a city area. The fuzzy model vehicle (FMV) is used for the MITRAM. We have tried to simulate the actual accelerator and brake operations to follow a leading vehicle by using the FMV constructed by neural networks. This FMV was trained with actual measurement data to reproduce the actual operations. The actual operations were reproduced in most of the simulation results. However, the simulation results greatly different from the actual operations also existed in some cases. It is thought that one certain operation was gained by experiencing many driving situations. Therefore, it was thought that some actual operation that we wanted to reproduce might not be reproduced by the FMV trained with only measurement data obtained during the same actual operation that we wanted to reproduce. In this paper, measurement data reference simulation was proposed as the technique of giving the training data which are not only the measurement data to the FMV. Since the FMV runs referring to the measurement data even when the FMV separates greatly with those in this simulation, ideal data which brings the FMV close to the measurement data can be made. We investigated whether some actual operation would be reproduced by the FMV which was trained with both data from the measurement data reference simulation and measurement data. It was found that the measurement data reference simulation was effective method as the technique of making the training data for the FMV.
日本シミュレーション学会の論文
2001-06-15

ファジィモデル自動車の再構成データを与える実測データ規範シミュレーションの提案と検討

スポンサーリンク

概要

著者

関連論文

スポンサーリンク