Research On Real Time Identification Method Of Urban Road Traffic State Based On Multi-source Data

As the gradually mature of the road network structure and further increase of the road traffic demand,urban road congestion is getting worse caused by the imbalance between the supply and demand of traffic space-time resources.The duration of the rush hour extended continuously,and the congestion range also extends from local to regional.Traffic congestion has been gradually transformed into normalized traffic phenomenon,which seriously reduces the safety and efficiency of road traffic.Therefore,detecting the recurrent urban congestion has gradually become one of the important part of the dynamic traffic management..However,the current traffic state identification methods are limited by the traffic parameters selection and the methods of traffic state estimation.The accuracy of traffic state prediction will improved if the space-time resources are used fully and the traffic conditions are identified promptly.In order to further improve the effect of traffic condition monitoring,this paper utilized weight distribution model to divided the standard of traffic parameters on the basis of improving the traffic parameter prediction model.Finally,using the cloud model to estimate the traffic state.This paper mainly includes the following five parts.The first part firstly describes the acquisition technology of dynamic traffic data and summarizes the common outlier detection method at the present.Then using the simple interpolation method and grey GM(1,N)model to repair the traffic abnormal data obtained by the detecting equipments.The second part aims at the phenomenon of the current travel time calculation model ignores the difference of each export direction of intersection.The paper puts forward the concept of upstream and downstream exit turning with the fixed detector data.So a new cumulative histogram model for real-time estimation of exit movement travel time on urban road established on the basis of the original one,which can effectively determine whether each export of the intersection is in traffic saturation by mixing the different detector data.The third part direct at the strong correlation between time and space of the traffic flow,which makes the target road easier to be affected by the upstream and downstream traffic flow.Afterwards,a model for short-term traffic volume forecasting was proposed based on the Kalman filtering model.The new model can unite the observed information and the metastasis of state variables,so that it is possible to obtain the optimal estimation of traffic flow.The fourth part taking into account the ambiguity and uncertainty that the traffic state exists,this paper selecting the membership function of normal cloud model to describe the fuzzy traffic state.finally,achieving the conversion between the qualitative and quantitative of the traffic state.The fifth part making use of the cloud model to analyze the traffic state in different time periods of the target rode based on the above research foundation.The practicability of the model is verify by comparing the actual traffic state.