Traffic Model Calibration: A Sensitivity Analysis

Abstract

Traffic-simulation modelling in urban networks is an increasingly popular tool for testing and analysing transportation problems. The modelling is useful mainly for overcoming the difficulties of real-world testing and offers the possibility of testing several alternatives in a safe and cost-effective way. However, the results obtained through simulation must be clear and appropriate to the actual traffic environment. Hence, for any simulation study, model calibration must be followed by validation to obtain the desired level of confidence. This research aims to contribute to the application of microscopic traffic models in urban networks. An area in the city of Porto was modelled using AIMSUN v5.0. Several analyses were performed to detect the sensitivity to single or multiple changes in the values of the models variables. The simulated results and a baseline model were compared to aid the selection and quantification of the different parameters for calibration. Furthermore, the importance of each variable in the traffic modelling process was assessed using tests and verifications to evaluate the models reliability. The complexity of the calibration can be significantly reduced because the impact of errors in the different model values on the accuracy of the results is known.