Lunch lecture: From Models to Machine Learning
Lunchlecture by Prof. Dr. Ir. Serge Hoogendoorn, Distinguished Professor Smart Urban Mobility
For decades, scientific research aimed at enhancing crowd safety and efficiency has relied on mathematical models to simulate pedestrian flow dynamics, aiming to predict and mitigate overcrowding risks. Unlike other areas in traffic engineering, crowd management has historically suffered from a scarcity of reliable and precise data.
However, the past decade has witnessed a transformative change, thanks to the introduction of various technologies for observing crowds. Although no single technology offers a complete overview, the focus of research has shifted from purely physical modeling to developing methods that use and integrate diverse data sources. This integration includes estimating and predicting crowding conditions, largely leveraging advancements in Machine Learning (ML).
This talk will begin with a brief historical overview of crowd dynamics modeling before delving into the latest in sensing technology and data science for identifying and predicting crowding risks. We will highlight several projects, including the ongoing Urban Mobility Digital Twin (UMDt) at TU Delft, that exemplify the application of different sensing technologies. Additionally, we will discuss recent advancements in using ML for short-term crowd management risk prediction and long-term forecasting for planning purposes. The presentation will conclude by exploring the ethical dilemmas posed by employing AI in crowd management.
Register now