Calibration And Validation Method Of Microscopic Traffic Simulation Model

Microscopic simulation has been widely applied to the research of traffic engineering, traffic planning, traffic management, traffic control and transportation design, due to its advantages on economy, convenience and experimentation. As an important step during the application of microscopic simulation, calibration of the simulation model can effectively ensure the accuracy of simulation and the reliability of the conclusions. The accuracy of microscopic simulation model needs not only the exact establishment of the route alignment and the traffic management and control measures, but also the reasonable expression of the characteristics for road users.Based on the characteristics of mixed traffic flow, the paper focuses on the calibration method of microscopic simulation model. Five of the most popular microscopic simulation software is reviewed, and the TSIS/CORSIM and VISSIM are specified, including the algorithms and the parameters. According to the two basic components of microscopic simulation model, transportation facility and driver-vehicle unit, the calibration framework can be divided into two parts, named as main program and calibration loop block. The main program proceeds by calling the calibration loop. The main program includes preparation of calibration, segment and bottleneck capacity calibration, calibration of intersection traffic volume, calibration of measures of system performance, post-assessment and model validation. The three phases, segment and bottleneck capacity calibration, calibration of intersection traffic volume and calibration of measures of system performance, calibrate the microscopic simulation model from different perspectives, and unify the whole and partial of the calibration of transport facilities. Calibration loop consists of calibration target establishment, evaluation of the default parameter values, sensitivity analysis and calibration of model parameters. The step of model parameters calibration consists of vehicle-type calibration and aggregate calibration. The vehicle-type calibration modifies the model parameters according to the differences between the traffic characteristics of each vehicle type in simulation and on real road, and the aggregate calibration modifies the model parameters according to the characteristics of macro traffic flow. During vehicle-type calibration, the paper puts forward a formula, called compound goodness of fit, to consider the different vehicle-type calibration errors, by assigning each vehicle type a weight composed of passenger car equivalent and the percent of the vehicle type. The weight expresses the contributions of different vehicle types to the traffic flow, especially the large vehicles. A freeway segment in Guangxi province is calibrated using CORSIM. Sensitivity analysis of different road units for each vehicle type and aggregate vehicles is carried out by establishing a new model which includes different types of road units, such as sharp curve, long steep slope, and short steep slope. The result shows that the goodness of fit for segment capacity is always within 2%, so there is no need to calibrate the segment capacity. During the calibration of vehicle-type travel time, the goodness of fit for cars, buses and each type of trucks are calibrated within 5%,5% and 15% separately, and the calculated compound goodness of fit is within 10%. During the calibration of aggregate travel times, the goodness of fit is calibrated within 5%. The result also proves that the calibration method is feasible.