Research On Monitoring And Forecasting Method Of Traffic State Based On Multi-source Data

efficiency of road, timely and accurate road traffic incident and traffic jams have been thefocus of research and difficult in the field of traffic condition monitoring. Moreover, withthe increase of traffic load, automatic congestion forecasting has become one of the hotspotsin the field of dynamic traffic management. There has sprung up a large number of resultsabout the forecasting of traffic congestion. However, due to the limitations of the traffic dataacquisition and processing methods, there are still large room to improve for the research oftraffic condition monitoring and forecasting in terms of efficiency, effectiveness andeconomy.In order to further improve the efficiency and effect of traffic remote monitoring andforecasting, under the guidance of the sharing principle for the related data, this paper hasconducted an innovative research on the multi-mode access to traffic data and traffic statemonitoring and forecasting based on the comprehensive utilization of multi-source data,including special vehicle detectors, inductive traffic control system, the road pricing system,vehicle tracking and positioning system and so on. The following is the mainly researchachievements.1）proposed a set of methods of traffic data acquisition and preprocess based onmulti-source dataIn regard to the weak basis of data for traffic monitoring and forecasting caused byserious shortage of traffic data and relatively backward of processing method, on the basisof the improvement of the processing methods for the data obtained from the existingspecial vehicle detectors, this paper proposed a method to obtain and preprocess the trafficdate based on the basic data of multiple business systems including the inductive trafficcontrol systems, the road pricing systems and vehicle tracking and positioning systems.Among them, for the traffic data obtained from special vehicle detectors, this paper hasdesigned a preprocess method of traffic data by comprehensively considering the single andcontinuous detections through further improving the preprocess flow of traffic data.In respect to the utilization of basic data from inductive traffic control system, usingSCATS as the representation, on the basis of the existing preprocess methods for the basic data, this paper proposed the concept of virtual time series of traffic data and the method toconstruct it, this paper also designed the preprocess methods of traffic data according to thecharacteristics of the time-vary of sampling cycle. In regard to the characteristics of roadpricing data, the preprocess methods of basic data was proposed, and it was a basis todesign the method of the acquisition and preprocess of traffic data obtained from roadpricing system. For the data characteristics of vehicle tracking and positioning system, thispaper designed the preprocess method of basic data obtained from GPS equipped floatingcar, as well as the method of data acquisition and preprocessing based on GPS data. Theresults of this study can provide a more economical, extensive and accurate data base fortraffic state monitoring and forecasting at a low cost.2）proposed a set of methods of automatic traffic incident detection based onmulti-source dataIn regard to the characteristics of traffic data obtained from four data sources includingspecial vehicles detectors, inductive traffic control systems, road pricing systems andvehicle tracking and positioning systems. The corresponding automatic incident detectionalgorithms were designed respectively. Among them, in the aspect of data source of specialvehicle detector, on the basis of the reasons analysis for the missed detection and falsedetection of the previous algorithms, this paper designed the online evaluation index systemfor the applicable conditions. Moreover, using the means of factor analysis and clusteranalysis, a new fusion method of automatic incident detection algorithm was designed. Inthe aspect of data source of inductive traffic control system, this paper proposed a newalgorithm of traffic incident detection by designing the combined variables based on trafficflow and average headway time. This algorithm was suitable for the road covered by thetraffic control system SCATS. In the aspect of data source of road pricing system, for thecharacteristics of data source acquired, this paper improved the standard deviation methodfrom three aspects including traffic fluctuations, the regularly traffic congestion as well asthe detection logic. A new algorithm of traffic incident detection for highway was designed.In the aspect of data source of vehicle tracking and positioning system, on the basis of theinstantaneous speed of vehicles, this paper designed a new algorithm of traffic incidentdetection for highway. In the aspect of the utilization of multi-source data, a new fusionmethod of automatic incident detection has been designed by analyzing the ability reflectingthe state of traffic incident of the various data sources. The results of this study can improvethe efficiency, effects and spatial coverage of automatic incident detection at low cost. 3） proposed a set of methods of automatic congestion detection based onmulti-source dataIn regard to the characteristics of traffic data obtained from four data sources includingspecial vehicles detectors, inductive traffic control systems, road pricing and vehicletracking and positioning system. The corresponding automatic congestion detectionalgorithms were designed respectively based on the connotation improvement of trafficcongestion index. Among them, in the aspect of data source of special vehicle detector, thispaper proposed an improved algorithm of the automatic traffic congestion detectionaccording to the regularity of traffic flow-speed curve. In the aspect of data source ofinductive traffic control system, this paper proposed a new algorithm of traffic congestiondetection according to traffic flow and average headway time. This algorithm was suitablefor the road covered by the traffic control system SCATS. In the aspect of data source ofroad pricing system, on the basis of the traffic flow and travel time data of path betweenO-D pairs, this paper designed a new algorithm of traffic congestion detection for highwayby comprehensively considering the natural links and extended links. In the aspect of datasource of vehicle tracking and positioning system, on the basis of the travel time of links,this paper designed an algorithm of traffic congestion detection for highway. In the aspect ofthe utilization of multi-source data, a new fusion method of automatic congestion detectionhas been designed. The results of this study can provide the space coverage and reliabilityof traffic congestion monitoring at a low cost.4）proposed a set of methods of multi-step forecasting of short-time trafficparameters and spatial and temporal scales of traffic congestion forecastingIn regard to the problem that forecasting effective is not good in the existing short-timemulti-step prediction of traffic parameters, it is caused by the fixed steps of forecasting.This paper designed a double-layer iterative model for multi-step prediction of trafficparameters based on the dynamic neural network, and also designed a double-layer for theshort-time multi-step prediction of traffic parameters based on the k-nearest neighbor, andthen analyzed the effect of both forecasts. For the current insufficient about trendsprediction of traffic congestion detection, on the basis of the result of the short-timemulti-step prediction of traffic parameters, this paper designed the prediction method forspatial diffusion range and duration time of traffic congestion. The results of this study canimprove the reliability of traffic congestion prediction, as well as further improve thereliability and predictability of decision-making by the traffic managers and travelers. The content, methodology and conclusion of this paper are all the improvement,complementary and useful exploration for the existing methods of the remote monitoringand prediction of traffic state. The outcomes of this study are of important academicsignificance and practical value to the further improvement for road safety and efficiency.