Intelligent driving diagnosis system applied to drivers modeling and high risk areas identification. An approach toward a real environment implementation

This paper considers the problem of characterize the way people drive applied to driver assistance systems without using direct driver signals. To make this, was developed an intelligent driving diagnosis system based on neural networks [2]. This take into account signals that can be acquired by a GPS data logging system: position, velocity, accelerations and steering angle. Now, are presented two approaches based on this intelligent driving diagnosis system. The first, propose to identify potential high risk areas on the road taking into account the average rate of diagnosis in each signal on the road. The second, present the structure of a driver model based on neural networks and using as inputs statistical transformations of the driving diagnosis time signals: inadequate steering and pedals inputs, speeding and road or lane departure. The validation of this model is developed in two applications: driver identification and driver classification (good or bad). In addition, an architecture based on fuzzy logic for real environment implementation is presented. System performance was tested in a driving simulation system [13]. Results presented in this paper shows that our intelligent driving diagnosis system allows to identify potential high risk locations on roads, and also is able to classify different kinds of drivers with a high degree of reliability.