Fabian Bastin (University of Montreal (Canada)) – Cinzia Cirillo (University of Maryland (USA))

Title: Large-scale transport model systems: from synthetic population generation to agent-based simulation

Abstract:

1. Generation of synthetic populations exploiting dependency structure transfer (Fabian Bastin)

The generation of synthetic populations consists of producing a realistic representation of a target population for modeling and behavioral simulation purposes, a task that can be tricky when the available multivariate samples are small. Starting from the idea that combining various sources of information makes it possible to design more realistic population generators, we introduce a new framework based on copulas to generate synthetic data for a population whose only empirical marginal distributions are known, using a sample from another population sharing a similar dependency structure. We compare the proposed scheme to common methods such as Iterative Proportional Fitting and to modern probabilistic approaches such as Bayesian networks, variational auto-encoders and generative adversarial networks. We illustrate on American Community Survey data that the method proposed allows to study the structure of the data at different geographical levels in a way that is robust to the peculiarities of the marginal distributions.

2. Large scale transportation model systems: an agent-based simulator for the State of Maryland (Cinzia Cirillo)

The talk will introduce the architecture of a large-scale microsimulation model that combines population synthesis, land use models, transportation choices, and agent-based transport simulation. This modeling suite overcomes limitations of aggregate four step models and is completely open source. Application of this type of tools will be illustrated in the context of transportation and equity; the first considers the problem of evacuating low-income carless populations during an emergency, the second evaluates the effects of complete streets on the willingness to walk and bike under different levels of traffic stress.

The seminar will take place in Room S08 at the Faculty of Sciences.