Research Of Calculation Methods For Urban Road Network Traffic State Based On GPS Of Taxi

With the increasing scale of modern cities,frequent traffic congestions,disordered traffic conditions and frequent traffic accidents have become the common problems of modern cities.Nowadays GPS is widely utilized for the detection of transportations.The taxi GPS systems have the advantages of shorter building time,low cost,wider coverage areas,highly accurate and in time information.Thus the supervisory and management of city traffic based on GPS taxis has been a top research topic of intelligent traffic systems.The GPS map matching algorithm of traditional probe vehicles is perfect enough but still can't meet the requirements of calculation of short-run large scale transportation in 5 to 10 minutes.Traditional matching algorithms often have defects of intensive and longtime computation.In our research,we utilize taxis as probe vehicles,and recognize the conditions city traffic conditions through the analysis of the data provided by taxis.After cleaning up the GPS data,we conduct the network matching through improved Markov model to improve the accuracy of matching.And then calculate the traffic condition of urban road network by the algorithm based on surface fitting.In the end,we can build a supervisory system of short-term city traffic.The works and points of creativity are as follows:(1)The quality of GPS-taxi data has direct effect on the accuracy of computed result of road traffic state.However,because of the terminal sampling equipment,internet delay,the effects of special weathers,it records lots of noise.In the essay,common noisy problems of GPS-taxi data are classified and generalized.Targeting at those different noisy problems,this article proposed appropriate data cleaning plans.(2)The geographic data of probe vehicles include only longitude and latitude data,which can't match the actual roads,and probably exists errors of several meters.Nowadays the mostly used method is nearest matching algorithm,nevertheless without guarantee of accuracy.This article proposed improvements for map matching algorithm based on implicit Markov model,and takes consideration of historical data to increase the accuracy in traditional short-run traffic condition algorithm.(3)Nowadays,how to use data samples to prognose the whole traffic condition is a hot research area.The traffic condition algorithm based on surface fitting takes advantages of superperformance compared with former research results.But surface fitting model based on the three functions doesn't support vector networks is susceptible to short-run accidental traffic conditions.In this article,RBF network is used to replace the algorithm of surface fitting to increase the robustness of traffic condition calculation.In the end,according to the real data of GPS taxis in Beijing,the algorithm of data cleaning proposed in this article can effectively eliminate GPS data to guarantee the quality of data.the algorithm of map matching,in comparison with formerly map matching calculation,has a higher degree of accuracy in short term forecast,and better performance on calculating of the traffic conditions.Those algorithms mentioned above have higher degree of creativity,efficiency and robustness and thus can meet the requirements of real time monitoring.