Research And Application Of Vehicle Location Prediction Algorithm Based On INS/GPS

Among the many applications of intelligent transportation systems,vehicle position prediction is particularly important.There are many researchs about vehicle location based INS/GPS integrated navigation system,whose goal is to provide more sustainable and stable navigation information when compared to standalone INS or GPS.When the GPS signal is briefly interrupted,the data can be integrated INS and GPS navigation continue for this data fusion algorithm.Although there have been several research works for fusing INS and GPS data to bridge navigation during GPS outages,most of them are offline methods and do not consider sensors data fluctuation,GPS signal interference or other issues which are likely to arise in extreme urban traffic scenes.Those scenes including traffic incident,inclement weather conditions or rush hour.Therefore,the vehicle position prediction algorithm is required to be more in line with real traffic scene,and can handle the resulting due to the increase in sensor error,modeling difficulties and so on.That is to say,the algorithm should be robust and stable in the practical application.This paper mainly studies on vehicle position prediction in ITS.During GPS outages,the algorithm can model and predict the vehicle position by fusioning INS sensor data.Then it can obtain continuous and reliable vehicle navigation information.The main works of this thesis are as follows:The characteristics and meaning of ITS are explained in detail.Makes the point that the vehicle positioning make an important role in modern transportation applications.Analysis of conventional vehicle positioning technology and some analysis of the features,advantages and disadvantages about INS/GPS integrated navigation system are given.For the data fusing issue in INS/GPS navigation system,there is an analysis about several existing fusion algorithm.Then by comparing their advantages and disadvantages,it can be drawn that the existing off-line algorithm can not adapt to the application of extreme urban traffic scenarios.A detailed analysis of some classical prediction algorithm in the field of vehicle position prediction is given.Analysis of the application infrastructure using in the vehicle position prediction field is given also.According to the characteristics of GPS and INS systems,some error sources that may arise during prediction are illustrated.Then this paper proposes a basic idea of the vehicle position prediction algorithm.A position prediction algorithm based on Online Support Vector Machine for Regression(OL-SVR)is proposed.The algorithm uses the online incremental predictive mode,and it can provide reliable vehicle position by using historical vehicle data to model and prediction.Simulation results prove that OL-SVR is more efficient and accurate in position prediction than PLSR and BPNN,achieving an accuracy improvement of 20.3%-64.8%.A position prediction algorithm based on Weighted Support Vector Machine for Regression(WSVR)is proposed,which combines a support-vector machine for regression with a weighed learning method.Because the relative significance of each point that might depend on the time difference between the points,the algorithm set into different weights for each training data based on the distance of data points and the prediction point.So it fill the blank that some incremental learning method assigned an equal weight to each time-series data point regardless of its relative order in the data set.Simulation results prove that WSVR is more efficient and accurate in position prediction than OL-SVR,achieving an accuracy improvement of 14.45%-62.52%.Using JAVA and MATLAB platform to achieves the data acquisition module,the algorithm module and the data visualization module.A simple and effective method for the implementation and evaluation of vehicular position prediction algorithm is achieved.