Research On Driving Intent Recognition And Lane Change Trajectory Planning Based On Aerial Dataset

Lane change is a common driving behavior,and if it is not performed properly,it may pose a threat to traffic safety and cause traffic congestion.Advanced driver assistance systems can provide safety prompts or automatically control vehicles to change lanes,avoiding lane change mistakes caused by driver negligence or fatigue,thereby improving road traffic safety.Currently,driver intention recognition and lane change trajectory planning are hot topics in lane change behavior research.With the continuous improvement of data collection technology,obtaining vehicle trajectory data has become more convenient,and many open-source high-precision vehicle trajectory data have emerged in recent years,providing a basis for studying lane change behavior.Based on this,this paper conducts research on lane change driving behavior on congested highways and expressways in China.First,the Aerial Dataset for China Congested Highway and Expressway(AD4CHE)is introduced,and then a set of rules and methods for extracting lane change behavior from aerial survey data is proposed.The Xgboost classification model is used to predict the driving intention of the lane change start point and decision point,and the prediction effects of the model at different lane change time nodes are compared.Furthermore,the paper analyzes the influence of different factors on driving behavior intention and classifies the driving styles of different lane change trajectory data using the K-means algorithm.The minimum tolerance threshold for various driving features of different driving styles is calculated based on the classification results.Based on clustering analysis and lane change time prediction,the optimal lane change trajectory is planned using a fifth-order polynomial with the minimum sum of lane change longitudinal displacement and driving stability weighted value as the optimization objective and the comfort evaluation indicator and safety evaluation indicator as constraints.A genetic algorithm is designed to solve the trajectory planning problem.The Prescan,Car Sim,and Matlab/Simulink simulation platforms are built,and the longitudinal and lateral control algorithms for vehicles are designed based on the vehicle two-degree-of-freedom dynamics model.Finally,three typical lane change scenarios are designed,and the planned lane change trajectories and the trajectory tracking control effects of the vehicle model under different driving styles are evaluated by analyzing the simulation experimental results under different scenarios.The experimental results show that the trajectory planning algorithm integrating driving styles proposed in this paper can achieve reasonable trajectory planning,can adapt to various typical scenarios,and can meet the needs of drivers with different driving styles.In addition,the lane change trajectory planning method proposed in this paper and the vehicle model used can ensure the comfort and safety during the lane change process.