Calibration Research And Implementation Of Macroscopic Traffic Flow Model Based On The Measured Data

Traffic flow theory is the application basis of intelligent transportation system,traffic management and control,traffic engineering design and so on.It is of great significance to enrich the study of traffic system.Most of the researches on traffic flow focus on the theoretical deduction and numerical calculation,but lack analysis and application of traffic data itself.In this paper,calibration method is proposed for the existing macroscopic traffic flow models with the measured traffic data,including the classical LWR model calibration method based on the least squares fitting method and the classical AR model calibration method based on the velocity item contrast method.After calibration,the model has better simulation effect under the actual road conditions.The rationality and validity of the calibration method are verified by the comparison of the numerical simulation results.Through the design of macroscopic traffic flow model calibration and simulation platform to achieve the above calibration method and verification process.Firstly,in order to provide the necessary data support for the calibration of the macroscopic traffic flow model,we study the processing and reconstruction methods of cross-sectional data and vehicle trajectory data which are widely used.For the crosssectional data,the calculation methods of detector micro-data,traffic flow,the average traffic speed are elaborated,especially the conversion between occupancy and density.As for the trajectory data,the kernel density estimation method based on Gaussian kernel function is proposed,and the boundary effect is reduced using modified reflection method.The NGSIM trajectory data is transformed into the macroscopic traffic volume successfully by this method,which achieves the ideal processing effect.Secondly,the first-order LWR model is calibrated and verified with the application of the measured data.By analyzing the theory and the measured data,the singleness of the Greenshield density-flow relation of the LWR model is proved.Therefore,we propose a fitting method of relation between density and flow based on least squares method.The results are expressed in polynomial form,and the calibration of the first order LWR model based on the measured data is realized.Meanwhile,by defining the evaluation index of the error and configuring the simulation input data,the calibration result of the LWR model is verified by numerical simulation.After calibration,the predicted errors of the LWR model with the three sets of experimental data are reduced by18.36%,12.78% and 8.75% respectively.Thirdly,the second-order AR model is calibrated and verified by applying the measured data.The pressure function of the AR model is hypothetical.The LWR model is transformed into velocity expression and is compared with the velocity term of the AR model.Then we obtain the pressure function which contains the density-flow relation that based on the measured data.The AR model calibration based on the measured data is realized after the above steps.The calibration results also support by the ARZ model theoretically.The calibration result of the AR model is verified by numerical simulation.After calibration,the predicted error of AR model with the three sets of experimental data are reduced by 45.10% ? 43.40% ? 49.79% respectively.Finally,the calibration and simulation platform for macroscopic traffic flow model is designed considering the repetition and complexity of the calibration and simulation process.The platform integrates several functions including traffic data loading,processing,model configuration,calibration,simulation,and comparison.The platform designed achieves the data processing,calibration and other methods proposed above to meet the different requirements of model calibration and numerical simulation.And the simulation results are finally shown.At the same time,a detailed functional demonstration of the platform is introduced.