Control-theoretic Driver model differentiation by linear and nonlinear input cue computation

Abstract

Context: To investigate the effectiveness of nonlinear calculation of the near and far angle for driver modelling in curve negotiation, simulated data is analyzed based of Mars driver model and the nonlinear realization of Mars model given in Scholtens. Aims: It is shown that due to the dynamic calculation of the near and far angle, the simulated driver is less sensitive to change of the Kp or Kc parameter value during a drive. Static calculation of the near and far angle showed a wider range of drivers that can be identified before the start of the drive, but is less adaptable to change during the drive. Methods: The range for the two parameters Kc and Kp was determined by setting two boundaries, whilst keeping the other parameters constant: the lateral position is bound to an absolute value of 1.8 meter and the steering wheel reversal rate to a maximum of 6. A ten by ten resolution range is subsequently established to form the basis for the simulations. The simulated results from the nonlinear and linear model are assessed on three criteria: realism, identifiability and descriptiveness. The three criteria show the realistic fit of the simulated data in comparison to real driver data. Results: Static calculation of the near and far angle can capture a broader set of human drivers front end if the parameter set of the driver is know front end. A dynamic calculation is more adaptable to an unknown driver set and is therefor in a smaller need of knowing the driver before the drive starts.