Lane Change Style Identification for Natural Driving Environments

In order to achieve accurate clustering and effective identification of driver lane changing style in natural driving environment, a driver lane changing style identification method for natural driving environment is proposed. Firstly, the improved hausdorff distance is used to perform spectral clustering and distinguish the lane change trajectory data. Nine parameters, such as vehicle transverse and longitudinal velocity and acceleration are selected to characterize the driver's lane-changing behavior. As well, the principal component analysis method is used to reduce the dimension of the characteristic data and synthesize the characteristic parameters that characterize the lane-changing style. Clustering and analyzing dimensionality-reduced feature data is then performed using the K-means, which is clustered out into three categories: aggressive lane changing style, cautious lane changing style and common lane changing style. Finally, the neural network is trained to realize the identification of lane changing style. According to the results derived from the identification of the HighD data set, the method can accurately detect the driver's lane-changing style, with an identification rate of 95.8%.