Civil Engineering Departmental Seminar - Modeling Large-Scale Pedestrian Networks: From Traffic Assignment to Machine Learning Approaches

Modeling Large-Scale Pedestrian Networks: From Traffic Assignment to Machine Learning Approaches

Understanding and modeling pedestrian networks in urban areas is crucial for effective infrastructure planning and sustainable transportation systems. This talk explores advanced methodologies for modeling large-scale pedestrian networks, synthesizing insights from user equilibrium pedestrian traffic assignment (UE-pTAP), dynamic pedestrian traffic assignment (DPTA), and machine learning approaches. We discuss a macroscopic UE-pTAP framework that addresses scalability issues by integrating pedestrian volume-delay functions (pVDFs) and stochastic walking travel times. Furthermore, we introduce a simulation-based DPTA model that incorporates bidirectional pedestrian streams, utilizing a dynamic user equilibrium (DUE) and link transmission model (LTM) to capture the unique microscopic characteristics of pedestrian flows. Additionally, we present empirical findings on pedestrian route choice preferences, emphasizing the impact of built environment factors such as navigational attributes, gradients, crossings, and Points of Interest (POIs). Leveraging mobile phone and crowdsourced data, we also examine the scalability of machine learning models for estimating walking and cycling volumes across super large networks, highlighting challenges in model training and inference. These insights enable data-driven optimization of pedestrian infrastructure, prioritizing investments to encourage walking and enhance urban mobility. This comprehensive approach offers valuable guidance for transportation modelers, policymakers, and urban planners in developing robust, scalable models for pedestrian network analysis