A Comparative Study On Traffic Flow Models Based On Measured Data

With the rapid development of China's economy, people's living standards continue improved, so the traffic conditions are put forward higher requirements. The motor vehicles are increasing sharply in some well developed cities. The increasing of the vehicles put up higher requirements of the road infrastructure. Some cities have built multi-lane freeways. Despite widening of the road and various other measures, the traffic problems are not solved, and have a more serious trend. Traffic congestion is a major inconvenience problem to people's daily life since people often waste a lot of precious time on this question. So, the research of traffic flow play more important role.The study of traffic flow can date back to last century, and academic achievements so far are fruitful. But most of the starting point of the study focused on theoretical analysis and numerical calculation. Traffic flow data collection and analysis of the work are ignored, and this paper attempts, using statistical data perspectives principle, to get the useful results for the traffic flow model and the model parameters, as are proposed based on the comparison and selection.Based on the project supported by NSFC and the DPIV method developed in experimental fluid dynamics, we recorded a lot of traffic videos and extracted many "headway-velocity" data from them using some computer programs. After the analysis of the data, we show that the distribution of headway and velocity belong to normal distribution, and the relationship between the average velocity and average headway correspond to the rule of traffic flow generally, and the average velocity of different lane changing rules is observed. All these show us the rationality of the data.Compared with Kerner's flow-density two-dimensional traffic flow model, we can find some data of high velocity with high density, and realize that as the samples' average density increases, the flow-density curvilinear of free flow that drew in the Kerner's model becomes weaker and nonlinear character becomes stronger. Beside that, after comparison study of manifold common traffic flow velocity-density models based on the above measured data, the superiorities of 1-d pipe-flow model is found in calculating the traffic flow parameters of expressway in our country. A basic method is proposed to draw the non-linear characteristics of traffic flow by 1-d pipe-flow model with changing parameters. Real measured data support and enlighten the study of traffic flow models. On the aspect of two-order macroscopic differential models, this paper propose a meth, based on lots of data's statistical character, to compare and study some well-known velocity-density relation functions and then choosing the best-fit function in each density section to form a piecewise function, which is much closer to China's present traffic conditions. On the aspect of cellular automata models, this paper give a meth to quantify the proportion of aggressive drivers, which is proposed by means of fitting different equilibrium functions to real measured data. On the aspect of MCLWR model, this paper give a rule to calculate vehicle's proportion of every level and give a meth to calculate velocity coefficient, which can be improved in MCLWR model as a important parameter.And the last, this paper discuss the traffic flow's unified theory of one-order and two-order differential models on the point of view of differential equations' first integral. Especially study on the 1-d pipe-flow model equations with general initial conditions, we can find an analytical solution, and construct an example to select the optimal computational format from different working conditions.