基于HighD数据集的高速公路小客车换道风险分析OA北大核心

To investigate the microscopic lane-changing behavior characteristics and associated risk for passenger cars on highways,we extracted lane-changing trajectories of the subject vehicle and driving data of surrounding vehicles from the HighD dataset.The focus is on the analyzing the subject vehicle's lane-changing behavioral patterns when there are vehicles ahead and behind in the target lane.Based on the lane-change conflict characteristics of light-duty vehicles on highways,we extracted lane-change risk indicators and established models for accident risk rates based on stopping distance coefficients and accident severity based on speed differences.Fault tree analysis was employed to calculate the lane-change risk index and evaluate lane-change safety.Additionally,the lane-change risk was classified into four levels using the k-means clustering algorithm.We established a lane-changing risk prediction model based on the gradient boosted tree(GBDT)framework to predict and validate the lane-changing risk for passenger car drivers using different feature combinations.The results show that the lane change risk prediction model with selected velocity-acceleration hybrid features provides the best prediction performance,contributing to improved lane-changing safety warnings for highway passenger cars.The findings can provide references for understanding the lane-changing behavior characteristics of passenger cars on highways,driving behavior pattern recognition and the parameter settings of driver assistance systems.