Crowd management in urban spaces: Developing a unified framework for simulating realistic human movement trajectories
Grant no.: NCN 2022/47/D/HS4/02576
Funding agency: National Science Centre (NCN), Poland
Funding scheme: SONATA
Funding period: 01.09.2023 – 31.08.2026 (36 months)
Budget: 366 080 PLN
Title in Polish: Zarządzanie tłumem w przestrzeni miejskiej: Opracowania ujednoliconych ram dla symulacji realistycznych trajektorii ruchu ludz
Research team:
Principal Investigator (Kierownik):
Investigators (Wykonawcy):
- Adriana Naumczuk
- Piotr Nyczka
– Ph.D. / M.Sc. / B.Sc. student
Collaborators (Współpracownicy):
- Julien Pettre (Rennes, France)
- Cecile Appert Rollan (Orsay, France)
- William H. Warren (Providence, USA)
Aims and scope:
Human civilization has experienced several fatal crowd disasters with the increased rate of urbanization. Overcrowding at massive social gatherings such as sporting events, music festivals, religious events, transportation facilities is a very common phenomena and is prone to life-threatening disasters. The death toll was 21 in Love Parade music festival in Duisburg, Germany in 2010, 3 during UEFA Champions League Final in Turin, Italy (with 1500+ injured) in 2021, 150 in Halloween celebrations in Seol, South Korea in 2022 and unfortunately the list goes on. The deadliest crowd accidents belong to the annual Hajj pilgrimage in Mecca and Mina, Saudi Arabia, killing 5000+ people in a span of about 25 years. Most of these incidents are consequences of poor crowd management by the organizers. To create effective measures of crowd management, one needs to understand the behavior of a crowd. The goal of this project is to create a combined model to explain multiple complex human-crowd situations emerging with specific patterns, e.g., formation of lanes are seen when two groups of people cross each other 180°. Our focus would be to combine the flows of pedestrian crowd without any spatial boundaries. We would analyze the self-organizing patterns of the crowd dynamics and use the results to calibrate and perform appropriate modeling of crowd behavior.
Tasks:
- Decompose the complex behavior of a moving crowd into an appropriate combination of inter-agent behaviors
- Determine the dependence of formation of spontaneously emerging patterns on crowd-density
- Construct a simulation approach that brings multiple complex scenarios of crowd movement under one common framework to generate realistic human trajectories
- Use deep neural networks to replicate the dynamic behavior of a crowd
Expected impact:
The findings of our research will upgrade the understanding of human-crowd behavior, which would be useful for crisis management in panic escape during emergency situations (fire, earthquake, appearance of a gunman etc.), as well as to improve pedestrian traffic management methods. The most important output of this research program would be a unified crowd simulation framework that can deal with a variety of complex crowd situations. This research has the potential to provide new ways to calibrate simulation techniques with real data – by making less assumptions about the underlying local interactions.
Publications:
Peer-reviewed articles in JCR-listed journals
2024 (1+), 2023 (0)
S. Worku, P. Mullick (2024) Detecting self-organising patterns in crowd motion: effect of optimisation algorithms, Journal of Mathematics in Industry (doi: 10.1186/s13362-024-00145-w).
Peer-reviewed articles in non JCR-listed journals
- …
Book chapters
2024 (1+), 2023 (0)
- P. Mullick, C. Appert-Roland, W. Warren, J. Pettre (2024) Methods of density estimation for pedestrians moving without a spatial boundary. In: Traffic and Granular Flow ’22 / eds. K. Ramachandra Rao, Armin Seyfried, Andreas Schadschneider. Singapore: Springer, cop. 2024. pp. 43-50. (doi: 10.1007/978-981-99-7976-9_6).
Conference papers
2024 (0+), 2023 (0)
- …
Forthcoming publications, submitted papers and work in progress
- P. Mullick, C. Appert-Roland, W. Warren, J. Pettre (2024) Eliminating bias in pedestrian density estimation: A voronoi cell perspective, submitted. Working paper version available from: (doi: 10.48550/arXiv.2408.03332).
