Analytical Method Of Traffic Condition Characteristics And Congestion Propagation Rules Based On Practical Measured Data

With the rapid development of urbanization,traffic congestion has become a common problem faced by many cities around the world.In order to effectively mitigate urban traffic congestion,intelligent transportation system(ITS)has been widely applied in metropolises.However,with the rapid development of the basic facilities of ITS,there are incongruities between the development of hardware and software of ITS,which result in the low intelligentize-based level of ITS.For a period in the future,traffic management departments will focus on how to extract useful knowledge and information from massive data,rather than how to obtain the traffic flow monitoring data.Therefore,traffic condition evolution and congestion propagation characteristics are studied based on measured data in this thesis,which can provide foundations for traffic management improving,infrastructure reform and travel guidance schemes generation.In this thesis,through summarizing and analyzing the existing research results,the spatio-temporal characteristics of urban road traffic flow data are analyzed firstly.And then,an improved spatial and temporal autocorrelation analytical method is proposed to study traffic condition and make a classification of it.Finally,On the basis of considering spatio-temporal non-stationarity of traffic flow,spatial and spatio-temporal models are constructed to describe the traffic flow of road network.And accordingly,the analytical methods of traffic congestion propagation are developed.The main work and conclusions are as follows.(1)From the angles of space and time,spatio-temporal autocorrelation,non-stationarity and periodicity of traffic data are analysed,and the capacity of traffic data to record traffic event is also studied accordingly.Through case studies of Beijing,it is found that there are significant autocorrelation of traffic flow in a certain range of space and time,and the autocorrelation decay with time and space.Furthermore,in the three aspects of traffic flow attribute value,the correlation between variables and the spatial autocorrelation,traffic flow presents obvious characteristic of non-stationarity.Because of the cyclical variation of traffic demands,traffic flow in the network is also periodic generally.The long-term cycle of traffic flow is 1 week and the short-term cycle is 1 day only.Because of the periodicity of traffic flow,the current disturbance in traffic flow data can effectively represente the occurrence traffic event,and through the computation of current disturbance,the severity and development of traffic event can also be revealed.(2)An improved spatio-temporal Moran's I index and Moran Scatterplot are proposed to study the autocorrelation of traffic flow and then make a classification of traffic condition.According to the improved spatio-temporal Moran Scatterplot,traffic condition is divided into 4 classes:homogenous uncongested traffic,heterogeneous uncongested traffic,homogenous congested traffic and heterogeneous congested traffic.Through the case studies of Beijing,the spatio-temporal autocorrelation of traffic flow is explored,and the spatio-temporal distribution and evolution of 4 kinds of traffic condition are also studied.There are significant spatio-temporal autocorrelation of the traffic flow in the road network of Beijing.There are higher spatio-temporal autocorrelation of traffic flow at weekends and holidays than at weekdays.This is mainly because travel demands decrease and get random in space-time at weekends and holidays.While in the day,the uncongested traffic of midnight and congested traffic of peak hours last long time and distributed widely,when trafic flow present high spatio-temporal autocorrelation.There was also some disciplinarians about 4 kinds of traffic condition in spatial and temporal distribution.In the time domain,the main type of traffic condition at midnight and peak hours are homogenous uncongested traffic and homogenous congested traffic respectively.While at noon,road sections with the heterogeneous traffic increase,because traffic demands become random in space-time.In space,homogenous congested traffic has a wide distribution on the space and time because of the aggregation of traffic demands in the urban centers.While with the location getting away from urban centers,as travel demands get sparser,the space-time range of homogenous congested traffic reduces,while the space-time range of homogenous uncongested traffic clusters increases.(3)Considering the spatial non-stationarity of network traffic flow,an improved SDM model is proposed to describe the spatial correlation structure of traffic flow.A method based on real data to analyze propagation characteristics of traffic congestion is then developed.Case studies on the road network of Beijing indicate that,traffic congestion on a local road section has the largest effect on the service level of all road sections in the network when the traffic density reach 22 pcu/km(at 14:00 p.m.).Generally,traffic congestion on a local road section may lead to the decline of traffic service level on surrounding road setions.With the increase of the space scale,the effect of local congestion on the other road sections recedes drastically.The effect on upstream road is greater than that on the downstream road,while the effect of local congestion on alternative road is the least.The key nodes with large effect on road sections in the road network locate around the urban centers and extend outside along the radial roads.This is mainly because that traffic flow on these key nodes and sections are all close to saturated state,so the traffic becomes sensitive to the disturbance of congestion degree.While in urban centers,as the traffic is with high density and low speed originally,congestion on local section has a less significant effect on the road network.(4)Considering the spatio-temporal non-stationarity and multicollinearity of traffic flow data,an improved PLS-STAR model is proposed to describe the spatio-temporal correlation structure of traffic flow.A novel method based on real data to analyze the spatio-temporal propagation characteristics of traffic congestion is then developed accordingly.Case studies on the road network of Beijing indicate that,the influence of local traffic congestion last 50 minutes,120 minutes,100 minutes and 110 minutes at 1:00,8:00,13:00 and 18:00,respectively.Thus,in the road network with severe traffic congestion,congestion propagation process lasts long and need a lot of time to get decayed,while congestion propagation process in uncongested road network lasts shot time and decays fast.During the day time,the road network is usually with high traffic density,so a large scale of network(within the eighth order neighborhood)is affected by local traffic congestion.However,at the night time when there is a light traffic,the propagation scope of traffic congestion becomes narrowed(only within the third order neighborhood).In detail,the duration of congestion propagation can be divided into two stage,propagation stage and convergence stage.In the road network with severe traffic congestion,the propagation stage of congestion may last more than 1 hour,and the convergence stage may be long too.While in the uncongested road network,propagation stage of congestion lasts only 30 minutes,and the convergence stage is usually short.