Adaptive Car-Following Behavior Identification by Unscented Particle Filtering

Several authors have argued that driving is in fact an adaptive process. That is, the behavior and the parameters describing this behavior in fact change over time and space, due to prevailing traffic conditions, road characteristics, weather and ambient conditions, etc. The empirical underpinning of these changes is however not a straightforward task. For one, this is due to the fact that until recently suitable data for these kinds of analyses were not commonly available. For two, it is caused by the lack of suitable dynamic parameter identification techniques. This paper puts forward a new approach to identify dynamically changing parameters of delayed car-following models, i.e. models that include a true reaction time. The approach is based on the unscented particle filter approach, which is generalized to enable estimation of the parameters of delayed systems (including the delay itself). The estimation of this true delay is achieved without linearization. Besides the methodological contribution, we show empirical evidence for changing driving behavior by applying the approach to real-life microscopic traffic data.