Research on vehicle lane keeping ability in low-speed scenarios based on real- vehicle driving behavior

Abstract In order to ensure the safety and stability of the vehicle lane in the low-speed scenario, this paper selects 2926 pieces of real vehicle driving lane maintenance in the low-speed scenario based on the open source natural driving data set AD4CHE in Dajiang, China, and extracts 15 typical driving safety and stability indicators in the low-speed scenario. Through K-means cluster correlation analysis, combined with Youden index, Boxplot method and statistical value, a total of five unreasonable indicators were eliminated, and a three-level evaluation system of lane keeping capacity under low-speed scenarios containing ten indicators was constructed, and the best discrimination threshold of each indicator was determined. The calculation method of subjective and objective weight combination based on difference coefficient is proposed to realize scientific and reasonable distribution of evaluation index weight. Finally, the evaluation model of lane keeping ability in low-speed scenarios based on multi-dimensional index fusion and the grading evaluation standard of vehicle lane keeping ability in low-speed scenarios are constructed. It can objectively and accurately reflect the actual situation of the driver's lane keeping ability in low-speed scenarios. Finally, the lane keeping ability evaluation model and standard in low-speed scenarios based on multi-dimensional index fusion are applied to the natural driving dataset, which verifies that the lane keeping ability evaluation model and standard in low-speed scenarios proposed in this paper are effective, and can provide theoretical support for traffic safety prevention and control in low-speed road scenarios.