Reproduction of stylized lane-changing scenarios based on natural driving data

Automobiles, as the best carriers for the integration and application of new technologies, are accelerating the transition towards intelligence and connectivity. Intelligent connected vehicles have become the strategic direction and competitive focus of the automotive industry's development. To construct a simulation scenario that allows intelligent vehicles to change lanes freely within heterogeneous traffic flows, this study builds a driving behavior data acquisition system to gather natural driving data characterized by multiple attributes. The free lane-changing behavior in multi-source heterogeneous data is extracted using a multi-constraint extraction method. Time-frequency domain features associated with free lane-changing are extracted and subjected to dimensionality reduction using Principal Component Analysis (PCA). Clustering of conservative, robust, and aggressive driving behaviors is then conducted using the K-means method. The trajectories of the ego vehicle and other traffic participants are accurately described based on spatial coordinate transformation and vehicle driving characteristics. Finally, a stylized lane-changing simulation scenario library based on natural driving data is constructed using Modeldesk software.