Study On Freeway Lane Change Decision Model Considering Driving Style

With the rapid development of highway construction in China,people’s travel is becoming more and more convenient,but at the same time,traffic safety is still a prominent issue.Although the traffic accident rate is declining year by year,the huge number of casualties and property losses are still shocking.Safety is an important issue in the field of transportation.In the future,as the Internet of Vehicles technology matures,the road smart infrastructure construction is steadily followed,and the autonomous driving technology achieves breakthroughs,the beautiful vision of Intelligent Vehicle Infrastructure Cooperative Systems(I-VICS)will also be realized.At that time,the requirements for vehicle driving safety will be more stringent,and correct driving decision-making is the prerequisite for achieving safe driving.Therefore,this article will conduct research on the lane-changing decision-making behavior:Through analyzing the NGSIM data,it is found that there are many outliers.By comparing the data processing technology,it is determined that the two-step trajectory reconstruction technology is adopted,that is,the wavelet transform is used to identify the outliers,the cubic spline interpolation is used to correct the abnormal points/blocks,the symmetric exponential average method and wavelet denoising method are used to denoise the horizontal and vertical trajectories.Through the definition of lane-changing behavior,the research focus and the principle of lane-changing trajectory extraction are clearly defined,and the trajectory data during a period of time before and after the lane change is intercepted from the complete trajectory.Through the identification of critical moments,the trajectory is divided into the car-following phase,the intention phase and the preparation phase,which are the model dependent variables.Based on the analysis of the decision-making mechanism of lane changing,12 factors that may affect the decision of lane changing were extracted.Perform variance analysis and correlation analysis on features to obtain features that have a significant impact on lane changing decisions.Considering the differences in driving behaviors of drivers with different driving styles at various stages,exploring the behavioral traits of drivers of different styles in the lane changing decision process,and analyzing the necessity of considering driving style factors into the lane changing decision model.The mean and variance of driving behavior parameters in different stages were compared,and the difference and reason of the parameters were analyzed.The Relief F method is used to rank the importance of the lane-changing decision features,and the key features of the lane-changing decision model are obtained: driving style,longitudinal speed of the self-vehicle,distance between the preceding and following vehicles in the target lane,speed gain,and speed difference of self-vehicle and preceding vehicle in the original lane.The lane-changing decision model is constructed based on the lane-changing decision-making characteristics and the lane-changing decision stage.Research shows that the effect of adding the driving style characteristic model to the model is increased by about 4%.The AUC of the XGBoost lane change decision model is as high as 96%,which is about 25%and 10% higher than the AUC of the SVM model and the RF model.The lane-changing decision model in this paper can be used in the on-board early warning system.Because the model takes into account driving style factors,the model can provide different drivers with more reliable early warning information in terms of time.On the one hand,it is beneficial to improve the driver’s acceptance and trust of early warning information.On the other hand,it can improve the service level of the decision-making assistance system,thereby improving driving safety.