Estimation Of Traffic Density In Urban Road Networks Based On Model Averaging

The estimation of traffic parameters of urban traffic network is a prerequisite for traffic state identification,road network capacity estimation and traffic guidance,which is also the focus of research in the field of traffic.In recent years,the traffic congestion in city has become more and more serious,and in response to this problem,the government proposes to unblock the traffic "capillaries".In this paper,based on the traffic flow theory and functional model to analyze the road properties of road network under the spatial and temporal characteristics,we propose a functional model-averaged traffic density estimation algorithm.The problem of estimating traffic density of intersections and road sections in urban road network areas is addressed.Firstly,this paper proposes Functional Linear Model(FLM)to estimate the traffic density on average by combining the characteristics of traffic data.The algorithm not only estimates the intersection and roadway traffic density,but also determines the intersection and roadway model weights.Secondly,it is demonstrated by numerical simulation that Leave-one-out Crossvalidation Model Averaging(LCVMA)and Fraction of Variance Explained(FVE)model averaging are applicable to different intersection traffic timing flows.While SIC(Smoothed Akaike Information Criterion)and SBIC(Smoothed Bayesian Information Criterion)are applicable to road sections,the road section model averaging methods are FVE-SAIC,FVE-SBIC,LCVMA-SAIC,LCVMA-SBIC.Finally,the practicality of the algorithm is verified by the example data of the urban section.It shows that the traffic jam does not occur on this road section,and the maximum capacity 2σ range is consistent with the actual situation.The problem of estimating regional traffic density of urban road networks is addressed.Firstly,this paper proposes a Functional Spatial Autoregressive(FSAR)model to estimate the regional traffic density on the basis of high frequency data.The algorithm estimates the regional average density from the perspective of spatio-temporal characteristics,and constructs the Macroscopic Fundamental Diagram(MFD)of the road network area to further measure the regional capacity.Secondly,the numerical simulation verifies that the FSAR model has a small mean square error and bias when the traffic data has different degrees of spatial dependence.The MMA(Mallows Model Averaging)method has less mean square error and associated model risk compared with SAIC and SBIC evaluation indexes.Finally,applying the algorithm to the regional data of the urban road network,the results show that the congestion in Guanshanhu area in the central city is more serious than other areas,and the capacity of this area can reach the maximum in the evening peak period.In contrast,there is a large gap between the capacity of the non-central city area in the morning and evening peak hours,and there is no traffic congestion in the non-central city area.In summary,from the perspective of urban roads,the proposed intersection and roadway FLM average estimation traffic density algorithm not only updates the traffic density in time and location,but also has more applicability and application prospects in practical applications.From the perspective of urban road network,the proposed FSAR model average estimation traffic density algorithm for road network area has theoretical basis and practical significance for traffic state identification and capacity estimation.