Modeling working networks (Pr. M. Saberi, UNSW, Australie)

While vehicular traffic networks have benefited from decades of advances in data collection and modelling methodologies, pedestrian networks remain limited by the lack of comprehensive infrastructure data, demand models, and network routing frameworks. This talk presents an integrated research program that addresses these challenges by developing the key components required for network-scale pedestrian modelling. The first challenge is constructing the walking network itself. A transformer-based deep learning framework is developed to extract sidewalks from aerial imagery, producing a routable representation of pedestrian infrastructure where no complete network previously existed. However, a network alone is insufficient without understanding how people perceive and use it. Street-level imagery is leveraged to estimate perceived walkability, providing scalable measures of the walking environment that capture qualities beyond conventional infrastructure inventories. Building on this foundation, ......