Publication

This paper concerns modeling of a driver's carfollowing behavior. The available signals, the distance to the vehicle in front as input and the driver's pedal positions, are not excited very much, and the recently suggested PrARX modeling framework is systematically investigated and compared with simpler model structures containing less tuning parameters. It is found that for long prediction horizons there is no gain having a more complex model structure than a linear ARX model. The one-step ahead prediction can be used for classication of whether the driver should be braking or not. Such a classier is shown to be better than a non driver-adaptive model in predicting when to start and stop braking.