Research on Identifying Social Preference Volatility of Vehicles behind the Target Lane

In mixed traffic flows, the volatility of the social preferences of vehicles behind the target lane (Vbtl) increases the complexity of lane-changing decisions for autonomous vehicles, thereby potentially raising the risk of traffic accidents. This study proposes a method for identifying social preferences based on key points. By analyzing steady-state and non-steady-state lane-change interaction scenarios from the NGSIM database, feature variables such as acceleration and jerk are selected. Principal Component Analysis (PCA) dimensionality reduction and K-means clustering algorithms are combined to classify social preferences. Additionally, a Support Vector Machine (SVM) classification model based on dynamic resampling and cost-sensitive learning is constructed. The identification of social preferences based on the Vbtl preference attributes for autonomous vehicle lane-change key points was successfully implemented on the Carla driving simulation platform.