Modeling of decelerating action in driver vehicle system

Driver control is an important consideration in the development of a vehicle driver-assist system. Using the driving simulator, this study gives consideration to a method for constructing a model of the driver's decelerating action with a Fuzzy Neural Network. Several kinds of headway distance and deceleration of the leading vehicle were set for the experiment. Measured values were fuzzy-clustered on the basis of the maximum deceleration and free running time. Input to the Fuzzy Neural Network was divided into learning and non-learning data by the holdout method, and the decelerating action was simulated using the model constructed for the non-learning data. It was shown that it is possible to construct a model reflecting driver performance using non-learning data and new data, and that fuzzy clustering of the input improves the precision of modeling.