Diagnosis methods applied to driver´s environment identification

Identifying the driver-vehicle-environment system can be compared to a diagnosis problem. To provide the diagnosis, some methods require an analytical or knowledge-based model of the system and others require only data-based model. Because no complete analytical or knowledge-based model of the system exist, methods of the second kind seem more appropriate. Among data-based methods, multiple correspondence analysis (MCA) has the needed features to reveal what are the most relevant variables that could better characterize the system. An example of MCA applied to characterize four driving situations is given. Prior to the analysis, recorded data is aggregated on 14 km. When this distance of aggregation is reduced at first no drastic changes appear in the quality of the identification. However, if the distance of aggregation is taken too short, the identification becomes deteriorated.