基于高相关目标驱动的车辆轨迹预测模型OA

Vehicle movement trajectory prediction is an important technology to promote the development of intelligent driv-ing.A vehicle trajectory prediction model based on attention is proposed.Firstly,aiming at the influence of different lanes on vehicle movement,an attention-based method for selecting highly correlated lanes and generating candidate goals is pro-posed,the method determines the highly correlated lanes by calculating the correlation score between the lane and map ele-ments,and then the candidate prediction trajectory endpoints are obtained by equidistant sampling on the highly correlated lanes.Secondly,according to the different degrees to which vehicle movement is affected by the surrounding environment,a method for calculating the correlation scores of candidate prediction trajectory endpoints based on branch attention is pro-posed,the method determines the correlation scores of the candidate prediction trajectory endpoints by calculating the atten-tion values of the candidate prediction trajectory endpoints and historical track,surrounding vehicles,highly correlated lanes,and vehicle lane interaction information,respectively.Finally,based on the correlation scores of candidate predicted trajectory endpoints,target selection is carried out to determine K predicted trajectory endpoints,which are then processed through a two-layer multi-layer perceptron to obtain the complete prediction trajectory.Experiments are carried out on Argov-erse dataset,and the results show that the proposed model is superior to the existing advanced models such as TNT and DenseTNT.