Key points are not available for this paper at this time.
Driving style plays a vital role in the personalized control of intelligent vehicles, and the current research methods based on supervised clustering have strong dependence on prior assumptions. This paper proposes a driving style semantic analysis method based on sparse inverse covariance clustering. This method introduces state continuity on the basis of Gaussian mixture model (GMM), and can extract the basic driving behavior units that are able to reflect driving style from multi-dimensional time series driving data without supervision, which is capable of realizing dynamic analysis of driving style. The NGSIM dataset is used to test the effectiveness of our proposed method. The experimental results show that the semantic analysis method can effectively identify and segment the feature switching between two continuous basic driving behavior units. Sparse inverse covariance clustering can effectively describe the characteristics of each basic driving behavior unit and the similarity of different driving styles. State continuity is introduced to extract basic driving behavior units from raw driving data, providing an interpretable semantic analysis method for dynamic analysis of driving styles.
Wei et al. (Fri,) studied this question.