Traffic model calibration, validation and statistical inference in a Bayesian framework

This thesis addresses the development and application of statistically rigorous methods for parameter estimation and inference in the context of traffic modelling. Fitting principle-driven mathematical models to observational data, in the presence of various sources of uncertainties and model imperfections, is a challenging statistical task. In this thesis, we adopted a Bayesian approach because of its flexibility and statistical coherence. We examined the inference problems and demonstrated the applicability of our methodology for several key traffic models including the carfollowing model, gap acceptance model, and a complex traffic intersection modelling software widely used by transport engineers.