Research And Implementation Of Vehicle Trajectory Prediction Algorithm Based On S-GAN

With the increasing of automobile ownership,the traffic safety problem brought by it is becoming more and more serious.Although the traditional vehicle safety control system can improve the vehicle driving safety to a certain extent,its role is still very limited because the human and envirommental factors in the process of vehicle driving are generally not considered Self-driving technology research has in the past ten years has made great progress,some self-driving cars have entered the testing phase,but as a result of the real traffic environment contains a variety of factors,self-driving cars on the road will inevitably interact with other vehicles,and how to analysis and forecast of these factors,it is study the challenge of self-driving vehicles in this paper,the research content as follows:(1)Put forward social generated against network was applied to vehicle track prediction in recent years,the generated against network have made good achievements in the field of pedestrian track prediction,thus inspired this paper it is introduced into the field of vehicle track prediction by LSTM network vehicle hidden feature extracting,repass pooling module calculation between the target and surrounding the vehicle position information by means of generating against network structure characteristics of global joint training back propagation of error,to obtain network parameters of each layer is reasonable,the decoding layer using has access to the context of the information generated forecast trajectory The experiment is carried out on the automatic driving data set,which proves the feasibility of the application of the social generated countermeasures network in the vehicle trajectory prediction field.(2)Against social generated against network interaction between vehicle features simple,does not take into account the driver behavior influence on vehicle trajectory,this paper puts forward a kind of social formation against network based driver intention recognition and track prediction method to extract the lane changing behavior intention recognition module connected features with the vehicles relative position,as the decoding of input layer,generated forecast track through the contrast experiment of multiple data sets,proves that the method is better than the existing methods in accuracy and real-time performance.(3)Based on the method proposed in this paper,a super parameter comparison experiment was conducted to analyze the influence of super parameters on the performance of the model.In addition,a vehicle trajectory prediction system was designed and implemented to compare the performance of different data sets and different algorithm models in an efficient and intuitive way.