Konferens: FAST-zero'15: 3rd International symposium on future active safety technology toward zero traffic accidents, 2015, Gothenburg

This paper proposes a method that uses a state-space model to predict vehicle behavior. It incorporates a proposed model of a generic driver, who is described by a simple feedback model with various constraints corresponding to general traffic rules and the physical limits of the vehicle. In addition, data assimilation is used to estimate the state and the model parameters on-line. By being able to predict natural driving data in unknown situations, the prediction error of the proposed model is significantly reduced compared to constant prediction, since data assimilation can be used to adapt the model to new situations. The data assimilation makes this model universal and robust against any new road-traffic situations. Furthermore, it is confirmed that this ability to predict can improve the fuel efficiency of adaptive cruise control systems.