Computational Driver Model in Transport Engineering: COSMODRIVE

Abstract

With regard to road safety issues, a deep understanding of the driver as a logic system is crucial to predict the most probable behavior according to the contextual elements. Knowledge and data about human functional abilities exist. But the problem is to organize and structure them. The development of a computational approach in driver modelization is addressed. In the first part, a brief historical overview is presented of available driver models in ergonomics and psychological areas, and the distinction between predictive and explicative models in an implementation perspective is the focus. In the second part, the computational aspect of the work is described, along with the software concepts, the cognitive modeling needs, and the implementation choices. Object-oriented techniques were chosen because they provide a modular overview of the general system and offer a convenient representation of cognitive processes. Object-oriented formalism, in particular object modeling technique diagrams, acts as a bridge between the two domains of computer science and the human sciences. The objective is to determine whether it is possible to implement reliably a driver model using the techniques from artificial intelligence and based on the theoretical knowledge from cognitive sciences research. This attempt to establish links between different scientific domains, requiring a common tool, is a challenge. A first step of a work that will have to be developed in a long-term time scale, taking into account its quite ambitious objective, is described.