PATH AVCSS researches have been traditionally oriented toward automatic vehicle design. Recently, the field of investigation has been extended from Automated Highway System (AHS) to assistance driving systems. One of the tools built at PATH for automatic system design and assessment is SmartAHS. SmartAHS is a micro-simulation tool dedicated to the simulation of automatic vehicles and has shown to be very useful for fully automatic system simulation. These simulations permit researchers to evaluate the impact of such system on throughput improvement. In order to apply the same method to the design of partially automated systems, it is necessary to develop a human driver component for this simulation tool. This component needs to allow the comparison of human driving characteristics versus automated vehicles in the long term, but first, it has to permit the production, for simulation purposes, of a realistic human driving behaviour. Modelling attempts of drivers behaviour in the field of Human Factors have been conducted since the early 60’s, in order to improve driving safety and driving learning. The first action taken was to describe driving behaviour by organising behaviour in various tasks, themselves part of more macro categories. The principal limit of this approach was to be only descriptive and almost not predictive. Moreover, most of the taxonomies were different and the limits between the categories were sometime confusing. The limits of this first trend motivated another approach, in which the models were based upon risk management. These models discussed risk evaluation, acceptation, and perception. These models introduced the notions of driver’s motivation, experience, and stressed the necessity to understand more about drivers’ cognitive activity, as risk perception and evaluation are strongly associated with the way a driver understands a driving situation. So, in the continuity of these efforts of modelling the driver’s behaviour, a new generation of models has recently been developed, emphasising the description of the driver process of thoughts. One advantage of these models is the potential to implement them, either by generating a simulation of the driver, or to integrate them to some already existing traffic simulation tools. Another application of this type of model is a direct integration either in the assistance system or in the design and development of the system. Actually, these models can be classified in relation of the aim of their conception. A possible classification is: i) the investigation of human reactions and learning of driving (e.g., The Generic Driver System (GIDS) Wierda & Aasman (1992)); ii) the design of invehicle devices (e.g., Integrated Driver Model (IDM) by Levison (1993) Allen’s model (1987); or iii) the improvement of the accuracy and realistic aspect of traffic simulation tools (e.g., ARCHISIM Espie & Saad (1995)). The goal of the modelling effort presented here is twofold. On one hand, there is an objective to design and evaluate AVCSS at a driver level (respective of human processing constraints), which imply a consideration of the cognitive processes involved while driving. On the other hand, there is also a goal to integrate of the model to a micro-simulation tool, for evaluation of AVCSS at traffic level, and more specifically in terms of throughput evolution. This second goal implies the consideration of vehicle models and control of the vehicle. SmartAHS mainly consists of automatically or semiautomatically controlled vehicles. This is why this component is called human driver model (as opposed to automated or semi-automated vehicles). The method preferred for the realisation of this model is a capitalisation of these various approaches by the application of a driver cognitive model, COSMODRIVE (COgnitive Simulation MOdel of the DRIVEr) (Bellet, 1998). This model conceptual framework is a skeleton around which can be organised the relevant aspects of the different approaches for the purpose of driver modelling. The general architecture of this model will be presented first, with a detailed description of the modules content and exchanges. In Section 2, the implemented modules and procedure of implementation will be described. The third section will be the description of the simulation realised with the model, for both normal driving and emergency case. Finally, this report will conclude with the description of a calibration procedure for part of the model. (Author/publisher)
Samenvatting