Research On Highway Lane Change Trajectory Prediction And Lane Change Risk Assessment Considering Vehicle Interaction Behavior

Lane changing is a common behaviour in motorway driving scenarios and can have an impact on the movement of vehicles in the vicinity of the changing vehicle,thus causing traffic flow disturbances.In addition,under high speed conditions,the reasonableness of the lane change can have a direct impact on road safety.However,existing research on vehicle lane change behaviour has focused on the historical trajectory information of the lane change vehicle itself,and consideration of the interaction behaviour of the lane change vehicle with the surrounding vehicles has yet to be improved.Therefore,it is important to investigate the interaction behaviour of lane change vehicles and the surrounding vehicles for lane change trajectory prediction and lane change risk assessment.Based on the above analysis,this paper focuses on the vehicle lane change behaviour and its interaction behaviour with surrounding vehicles,constructs a lane change trajectory prediction model considering vehicle interaction behaviour,establishes a quantitative assessment method for lane change risk and carries out lane change risk assessment on the predicted trajectory,in order to achieve the purpose of accurately describing the lane change process and providing new ideas for lane change safety warning.The research content of the article mainly includes the following aspects:(1)Analysis of vehicle lane change characteristics and vehicle interaction characteristics.Based on the lane change trajectories extracted from the German High D dataset after preprocessing,the lane change process is divided into phases and the basic theory of vehicle interaction behaviour is described.The motion characteristics of the vehicles in each stage of the lane change are analysed to understand the motion mechanism of the vehicles in the lane change from a statistical perspective.Spearman correlation analysis is used to investigate the interaction between the lane change vehicles and the surrounding vehicles.The results show that there is a correlation between the motion state of the lane change vehicle and the motion characteristics and relative motion indicators of the vehicle in front of the target lane,the vehicle behind the target lane and the vehicle in front of the original lane.(2)Establishing a prediction model of lane change trajectory considering interaction behaviour.Considering the interaction characteristics of vehicles,the interaction relationship between vehicles is regarded as the "social behaviour" between vehicles,and a social-LSTM model based on the prediction model of vehicle lane change trajectory is established.The Social-LSTM model was trained using the lane change trajectories of 8710 vehicles extracted from the High D dataset,and the final model structure consisted of a 3-layer LSTM model and a fully connected layer.The deviation of the predicted points from the actual points is small,and the predicted trajectory will fluctuate in the later stage.The predicted trajectory fits better in the early stage compared with the actual trajectory,and will fluctuate slightly in the later stage,and the maximum deviation value from the actual trajectory is around 0.3m,which only accounts for 7.5% of the total lateral offset of the vehicles,and the trajectory prediction model has a good effect.(3)Evaluate the risk of lane change for the predicted trajectory.Based on the conflict form of lane change vehicles,the dynamic change trend of conflict points between lane change vehicles and surrounding vehicles was analysed,alternative safety indicators SDI and TDTC were selected as conflict indicators of lane change vehicles,and then the conflict risk exposure level REL and risk severity RSL of lane change vehicles were calculated based on SDI and TDTC,and the accident tree method was used to determine the system risk.K-means was used to classify the risk level of the lane change system into 4 categories.The system risk is finally calculated using the predicted lane change trajectory data,and the system risk is assessed and predicted for the lane change trajectory.In summary,this paper mainly considers the interaction behaviour of lane change vehicles and surrounding vehicles,analyses the interaction between lane change vehicles and surrounding vehicles,establishes a prediction model for lane change trajectories considering vehicle interaction behaviour based on the Social-LSTM model,and carries out lane change safety evaluation on the basis of the predicted trajectories.The research results of this paper further explore the characteristics of the interaction behaviour of lane-changing vehicles and surrounding vehicles,which can enrich and improve the research results of existing vehicle lane-changing safety warning models,provide new ideas for vehicle lane-changing safety warning from the trajectory level,provide a way to effectively analyse and predict the complex interaction behaviour and risk level on highways,and in practice can provide a reference for highway management departments to risk assessment and control tool.