Calculation Of The Traffic State Map

As the further process of urbanization of the society, the urban trafficnetwork congestion is increasingly worse. As a result, the demand of usingIntelligent Transportation Systems to monitor and manage the city road net-work is becoming more urgent. As the main research field of IntelligentTransportation Systems, traffic state estimation can calculate real-time trafficstate in the road network depending on the sensor data, which is what trafficflow analysis based on. In addition, due to the size and complexity of the ur-ban road network, the entire road network unified management and control isnot realistic. So the partition of the urban road network has been more im-portant. In this paper, two different sources of traffic data, GPS data andSCATS data, are used for traffic flow state estimation, and a method to parti-tion the urban road network dynamically based on traffic state calculatedpreviously over a period of time is proposed.The GPS system is able to provide the location, the direction as well asthe speed of the car, which is easily to be collected and has a wide range cov-erage of the map. Therefore, it becomes one of the main research area of traf-fic state estimation based on the analysis of car GPS data. In this paper, weresearch on the road link. Using the speed’s continuity of time and space inthe internal of a road link, we use the cars on the road link as sampling pointsto model the traffic state by surface fitting method. As the basis of the analy- sis, we have also introduced the conversion of GPS data and map matchingmethod.SCATS system as a traffic system widely used currently, of which muchinfrastructure has put into use. In order to fully utilize the SCATS system, theanalysis of SCATS Loop data has also gradually been studied. As a clearmathematical relationship does not exist between the SCATS Loop data andthe actual traffic flow speed, we use the methods of polynomial fitting, sup-port vector regression and BP neural network for supervised training to getthe relationship and a comprehensive comparison among these methods aregiven.For these two methods of traffic state estimation, experiments are bothcarried out based on the real speed value obtained from video artificial cali-bration in some road link of Shanghai. And we analyze and verify the accu-racy of the results of the traffic state estimation from the experiments.Finally, according to the results of the traffic state analysis, as GPS da-ta’s wide coverage, its estimated traffic status is used as the input of the dy-namic partition of the road network. By sampling on the whole road network,the map is divided into regions of same size. We cluster the map regions bylocation of each grid and traffic states over a period of time, then obtain theresult of partition. On the partition basis, according to the traffic state ofSCATS, we propose a regional traffic index, and analyze region’s characteris-tics of the road network with the traffic index.