A simple realistic driver model

Many of the published driver models concentrate on algorithms which achieve accurate path following at the expense of realistic replication of the driving task itself. In this paper we consider only the most basic driving responses, to achieve a simple yet surprisingly realistic driver model. Lateral control is based on steering corrections aimed at projecting the vehicle onto a path at a single preview point on the road ahead. Only the preview time and a single proportional gain are required parameters, supported by a basic approximation of understeer gradient which becomes proressively more important as desired lateral acceleration increases. The longitudinal model regulates speed solely in proportion to an estimate of the length of road the driver can see ahead. Both aspects of the model are executed in a computationally efficient way, using the simplest possible definition of a track. The model is tested for robustness in simulation, and it gives intuitively sensible responses. Results are then given in comparison to vehicle tests, with the longitudinal parameters tuned to match the measured driving behaviour of two test subjects, while nominal lateral parameters are shown to be effective. Finally, the model is also shown to be capable of reasonable, if approximate prediction of speed variations for the same test drivers on an independent section of road.