Driver´s environment identification using automatic classification methods. Active safety application

This paper presents an application of diagnosis methods to the driver-vehicle-environment system. The study aims at characterizing the driving environment on the basis of data collected on the vehicle. 14 candidates drove on a driving simulator through 4 different driving situations: driving on a motorway with a dense traffic, driving on an A road with a dense traffic, driving on a motorway with a light traffic and driving on an A road with a light traffic. Multiple correspondence analysis (MCA) was used because it could provide a diagnosis without requiring a model of the studied system. Because MCA doesn´t allow automated data classification, this first analysis is followed by a discriminant analysis, which provides 97% of well diagnosed driving situations. Finally, prospects that could enhance the diagnosis reliability are exposed.