Parameter Identification of Stylized Free-lane-changing Bezier Model Based on Behavioral Clustering

Abstract

In order to realize the identification of driving style differences in intelligent vehicle free lane changing decision in heterogeneous traffic flow, a driving behavior data collection system was built in this research to obtain the natural driving data of multi-attribute drivers, and the free lane changing behaviors of heterogeneous data were mined based on multi-constraint extraction method. The time frequency characteristic parameters associated with free lane changing were extracted, and the conservative, robust and radical behavior clustering were completed based on PCA and K-means methods. The bezier curve was selected for modeling, and the stylized behavior data were input based on the genetic algorithm for parameters identification. Finally, three types of stylized free lane changing models were obtained. The stylized identification models were evaluated by random lane changing behavior data. The RMSPE value of the model can be obtained, and the prediction accuracy of stylized bezier curve was better than the that of the random lane change curve.