A fundamental challenge in constructing Big Models is the question of calibration. Big Models typically have a large number of parameters, which need to be inferred from data in a robust and theoretically justifiable way. Approximate Bayesian Computation is one promising approach to tackle this calibration problem for large complex models.
ACEMS CI Tim Garoni and AI Lele Zhang have been working with VicRoads over the past several years to develop a flexible and computationally fast stochastic model of urban arterial road networks. In this project, they team with ACEMS CI Scott Sisson to apply novel techniques from the field of Approximate Bayesian Computation in order to tackle the calibration problem for this large-scale traffic model.