This paper proposes a structure for an “active” model of driver that enables to predict behaviour and performances in dynamic changing traffic conditions, with potential application both offline and online. A simple prototype of the system has been realised in software, and has been compared against observed data in a rudimentary validation. The comparison reveals that the software's outputs accord reasonably with the observed values, not only in terms of central tendency but also in terms of capability to predict the between-driver variability. The next step is to create a system capable of identifying driver characteristics and state from observed data. However, further research is needed in order to expand the model in several dimensions, primarily to represent more complex scenarios in the presence of advanced automation technologies. (Author/publisher) Reprinted with permission from Elsevier.
Samenvatting