It is important to develop the driving assistance system that conforms to the individual driver’s characteristics so that the driving assistance technology is widely accepted. Timing and ways of the assistances should fit in individual characteristics of unusual driving behavior.

In order to measure usual driving behavior, we developed test vehicles with an on-board driving recorder and various types of sensors. Several driving parameters (e.g. the driver’s operational measures, the vehicle state measures, the traffic condition measures) were recorded with the recorder.

The equipped test vehicles were given to 67 paid volunteer drivers to use for two months in commuting situation. The field operational test was conducted a total of about 1800 driving trips. The measured operation behavior was accumulated to a database.

In order to develop a new driving assist method for the stopping maneuver, we modeled the driving behavior for usual driving by the Bayesian network which was a graphical model obtained from statistical analysis of the measured behavior data. A deviation from the usual operation behavior (i.e. unusual behavior) can be estimated using this model. The unusual behavior can be regarded inappropriate behavior and thus can be applied to a cue for the driving assist such as warning presentation.