Vehicle Location Awareness And Trajectory Prediction Algorithm In Complex Urban Environment

Vehicle positioning technology brings convenience to the continuous acquisition of large-scale accurate trajectories.In today's intelligent transportation systems,large-scale vehicle trajectory data has high utilization value.In most cases,the widely used GPS positioning technology can accurately provide real-time vehicle position information.However,in some relatively extreme traffic environments,such as tall buildings sheltering,commuting to and from work,elevated turns,etc.,GPS positioning cannot continuously provide reliable and accurate vehicles location information,which has a great impact on people's safe travel and the development of intelligent transportation systems.Therefore,how to obtain reliable vehicle location information in this complex urban traffic environment is an urgent problem to be solved.This paper implements multi-source data fusion based on GPS/OBD integrated navigation system to study vehicle location awareness and trajectory prediction algorithm in complex urban traffic environment.The main work of this paper includes:Firstly,combined with the characteristics of intelligent transportation system,a new integrated navigation and positioning system,namely GPS/OBD integrated navigation multi-source information fusion system,is proposed.By analyzing the characteristics of GPS/OBD integrated navigation system,the sources of error in vehicle position prediction are described,including coordinate conversion error,data acquisition error,and sensor error.And based on GPS/OBD,several traditional vehicle position prediction algorithms are analysised,and their advantages and disadvantages are pointed out,then it explains that the use of existing complex traffic environment has certain limitations.Secondly,aiming at the shortcomings of existing prediction algorithms,this paper proposes a vehicle trajectory prediction method based on GPS/OBD integrated navigation(LS-PSO-SVR),which uses support vector regression to train and model the historical trajectory data of vehicles.At the same time,the local shrinkage part icle swarm optimization algorithm is used to optimize the parameters of SVR,and a reliable and effective prediction algorithm model is obtained.At the same time,combined with experiments,the accuracy of the algorithm is improved by 40.5%-95.5% compared with BPNN,GPR and PLSR.Finally,for the existing algorithms,the error accumulation occurs in the late stage of GPS failure,and the location information of the vehicle in the GPS failure phase is closely related to the adjacent historical trajectory.T his paper proposes a bidirectional prediction model based on KRR.Therefore,KRR is selected as the basic prediction method,and bidirectional prediction is specifically used to construct the prediction model.Thus it can improve the accuracy of vehicle po sition prediction and achieve the vehicle trajectory integrity.Finally,it is proved by experiments that the accuracy of this algorithm based on KRR is improved by 14.63%-63.33% compared with LS-PSO-SVR.This paper uses Python and Java language programming,and realizes data processing and algorithm implementation through Matlab platform.Finally,the predicted data is visualized by Google Earth.This provides a simple and effective method for the implementation and evaluation of vehicle podition awareness and trajectory prediction algorithms.