Driver-Vehicle Interaction : Identification, Characterization and Modelling of Path Tracking Skill

Abstract: Since the dawn of the automobile, driver behaviour has been an issue. Driving can result in accidents that may harm not only the driver but also passengers and the surroundings. This calls for measures that restrict the usage of vehicles and to assist the individual driver to conduct the driving in a safe, yet practically efficient manner. The vehicles should therefore be both safe and intuitive, and preferably answer to thedifferent needs of all kinds of drivers.Driving skill can be defined in many ways, depending on the objective of the driving task, but answer in some way to the question of how well the driver can conduct the driving task. To assist low skill drivers without compromising the driving demand for high skill drivers, it is of highest importance that vehicles are tested and designed to meet those needs. This includes both the testing activities in the vehicle design phase in general but also the configuration for active systems and preventive safety, preferable with settings that adapts to the skill of the individual driver.The work here comprises the definition of skill and of driver recruitment procedures, scenario design, the development of an analysis method for objective measures, and the gathering of metrics to characterize the driver skill. Moreover, a driver model has been developed that makes use of driver skill characteristics. To gather the information needed, extensive multidisciplinary literature studies were conducted, as well as using field tests and test using an advanced moving base driving simulator. Here the focus is on path tracking skill, which is the main control aspect of driving, although the developed driving scenarios allow a varying degree of path planning, which is more related to regulation. The first simulator test was done with a very simple criterion fordriver selection, but the results gave a good insight into the variation between drivers ingeneral. For the following tests the recruitment procedure was refined to find drivers with high or low vehicle control and regulation skill, a recruitment that also was verified to really represent two different populations.A method was defined that successfully identified sets of skill-related measures, with some variation in composition depending on the path tracking demand on the driver. Int he curving road scenario, for example, the highest number of skill-related measures is identified in the curves, which is reasonable since the straight segments do not require the same amount of active control from the drivers.The driver model developed uses a quasi-static analytical description of the driver knowledge of the vehicle dynamics, but possesses the capability of nonlinear descriptions. The parameters in this model are mainly physical properties that easily can be related to the driving process. Metrics gathered are used for identification of the driver model setup for a double lane change scenario using an optimization routine, with adjusted parameter settings for different velocities.With a subjective comparison of the recorded driving simulator data, the method is verified to enable driver skill settings for driver models. In addition, the method allows metrics to be gathered for driver skill identification routines, meeting the defined objectives of the project.