The Research On Interrupted Traffic Flow-based Speed-density Model And Short Term Prediction Of Traffic State

Through the integration, analysis and mining of all kinds of traffic data, intelligent transportation system makes traffic management to be more accurate, more efficient and more timely, where traffic data analysis is the basis of the system. As one of traffic fundamental diagrams, speed-density relationship can provide a solid foundation for traffic flow analysis and efficient traffic management. Because of the change of modern travel mode, with the dramatic increase in the number of vehicles in urban, traffic density increases similarly, and due to the impact of traffic signal and other factors, vehicles change velocity frequently, which result in that the speed-density model based on uninterrupted traffic flow is not applicable for interrupted traffic flow. What's more, based on information entropy theory, comparing to a single traffic parameter, the speed-density relationship contains more abundant traffic information, but achieve traffic flow prediction, and existing researches for short term traffic prediction are all forecasting single traffic parameter.Based on the analysis of a large number of coil data of urban road, a new method which can accurately describe the speed-density relation of the interrupted traffic flow is proposed for speed fluctuation characteristics. This method divides speed data into small data set meeting the normal distribution, obtains upper quantile and lower quantile of normal distribution and gets upper and lower quantile sets of all normal distributions, then fits two quantile sets to get upper curve and lower curve to descript the supremum and the infimum of speed. In addition speed, density and occupancy are united to predict short-term traffic by Markov model. The values of speed, density and occupancy are respectively divided into five, four and three traffic state intervals by existing research results to reduce the mutation characteristics of traffic data. Then speed, density and occupancy are united to generate four union parameter states including(speed, density),(speed, occupancy),(density, occupancy) and(speed, density, occupancy) as the state sets of Markov model, and comparative analysis the prediction accuracy with single traffic parameter.Variant models are achieved for the outer lane and the inner lane by fitting massive coil data. The speed-density relation of the outer lane is in accordance with the logarithmic model, while the speed-density relation of the inner lane meets a segment model instead of a single model, where when the density is less than critical density, it conforms to exponential model, otherwise logarithmic model. The divergence of two verifies that the characteristics of traffic flow in the outer and inner lanes are different. When predicting short-term traffic state, among all traffic parameter combinations, the united state of(speed, density, occupancy) has the lowest prediction error rate, single traffic parameter has the worst performance, and the prediction effect of united states of two traffic parameters are between them. So the results show that the proposed speed-density model can accurately and intuitively describe the state of urban road traffic, and united parameters has a better prediction performance compared to single parameter.