Abstract:At present, the differences between drivers are seldom considered in decision-making and planning of autonomous vehicles. This paper proposes a method for extracting personalized driving indicators for drivers in specific scenarios and generating a planner that accounts for driving styles based on these indicators. A fuzzy clustering method with temporal constraints is used to classify the experimental data. The statistical sample test method is introduced to analyze the significance of the difference of each variable data, and the variable indicators that can best reflect the personalized driving characteristics of drivers in typical driving scenarios are extracted. Finally, the parameters of the Artificial Potential Field model are calibrated by combining the driver's personalized characteristic variables. Taking the car-following scene as an example, the calibrated car-following model is utilized to plan the car-following trajectory, in which reflects different personalized characteristics.

Analysis of Naturalistic Driving Behavior for Personalized Autonomous Vehicle