Crowd dynamics and evacuation research
Department of Mathematics and Systems Analysis
"Hotspots", or crowded public places, pose safety and evacuation challenges. Our crowd dynamics and evacuation research focuses on studying issues related to these socio-physical systems.
Crowd evacuation research - history

Our crowd dynamics and evacuation research started in 2005 when Harri Ehtamo, Simo Heliövaara, Simo Hostikka, and Timo Korhonen started collaboration between the Department of Mathematics and Systems Analysis at Aalto University and the Fire Safety Research group at VTT Technical Research Centre of Finland. Together with VTT, we have been developing the widely-used fire and crowd dynamics simulator FDS, and an evacuation simulation module Evac of it (Korhonen and Hostikka, Technical Report VTT, 2009). FDS+Evac has been used, e.g., in the fire safety design of Helsinki Music Centre. The research collaboration has also resulted in three Ph.D. theses during the years 2005-2021.

One of the highlights was the doctoral dissertation by Anton von Schantz that took place in Aalto-University in June 2021, with a related one page science news article in the newspaper Helsingin Sanomat (Helsingin Sanomat/Tiede, 24.6.2021, p. 79).

Research group
Harri Ehtamo
Aalto University School of Science
Simo Heliövaara
Nordic Investment Bank
Anton von Schantz
Aalto University School of Science
Simo Hostikka
Aalto University School of Engineering
Timo Korhonen
VTT Technical Research Centre of Finland
Juha-Matti Kuusinen
KONE Technology and Innovation
Jaan Tollander de Balsch
Aalto University School of Science
Bottleneck phenomena of crowd congestion

Crowd dynamics describes pedestrians moving in crowds along pathways and walking areas in market places, traffic stations, etc. Often pedestrians walk in a coordinated manner, and they are well aware of their goal of walk.

corona Large mass gatherings and crowded festivals may cause crowd congestions. Sometimes, the threat to the crowd members is the escaping crowd itself that may originate panic. If there are too many pedestrians in a scarce or dwindling space, a stampede or escape panic may occur. According to empirical investigations: (1) People move or tend to move considerably faster than normal. (2) Individuals start pushing, and interactions among people become physical in nature. (3) Moving and, in particular, passing of a bottleneck becomes uncoordinated. (4) At exits, arching and clogging are observed. (5) Jams build up. (6) The physical interactions in the jammed crowds add up and cause dangerous pressures, which can push down brick walls (Helbing et al., Nature, 2000).

These collective phenomena of escape panic can be modeled in the framework of physics-based computational evacuation models, such as the dynamical social force model developed by Helbing et al.

Counterflow effects

We present a model for agents' behavior in counterflow situations, where they try to avoid collisions with other oncoming agents. Numerical simulations with the model are experimentally validated. Correct modeling of counterflow is important because counterflows are known to slow down the evacuation and even cause mutual blockages (Heliövaara et al., Building and Environment, 2012).

Evacuation experiments

In our experimental study, students evacuated from a corridor under a cooperative, and a competitive evacuation pattern. In the former, they were instructed to minimize the whole group's evacuation time. Whereas, in the latter, they were instructed to minimize their own evacuation time. Based on the video material, several behavioral phenomena were observed. Compared to other similar experiments presented in the literature, a surprising observation was that evacuation is significantly faster under the selfish evacuation pattern (Heliövaara et al., Safety Science, 2012).

Evolutionary evacuation game

FDS+Evac For a competitive exit congestion, we develop an evolutionary game, where the agents play against their nearest neighbors. The agents have two strategies to choose from: Patient and Impatient. It is assumed that these strategies correspond to their patient and impatient behaviors in an actual play of the game. In a numerical congestion simulation, the proportion of impatient agents increases with the distance to the exit. This is natural since here time is a limited resource.

In the figure, along any semicircle centered at the exit, the proportion of impatient agents approximates the game's evolutionary stable strategy (ESS) (Heliövaara et al., Phys. Rev. E, 2013; von Schantz and Ehtamo, Phys. Rev. E, 2015; Physica A, 2019).

Non-monotonous crowd dynamics

The evolutionary congestion game coupled with a computational evacuation model results in non-monotonous crowd speed and pressure patterns, the green and violet halos in the figure. This is because impatient agents in the back of the crowd overtake the agents in front of them. The computational results coincide with the above experiment. In particular, in a competitive evacuation pattern, faster students overtake their predecessors, and in a cooperative pattern, i.e., all students are patient, they stick to their positions within the crowd throughout the evacuation (von Schantz and Ehtamo, Physica A, 2019).

Optimization models for crowd evacuation using rescue guides

In addition to models being able to describe harmful crowd phenomena, they should also prescribe solutions to prevent them. We develop new mathematical models and algorithms to solve the minimum time crowd evacuation problem with rescue guides. The new methods are based on mathematical optimization, namely scenario optimization, genetic algorithms, numerical simulation-based optimization, and bi-objective optimization. The methods are applied to the evacuation of a concert venue. The solution gives the number of guides, their initial locations, and exit assignments that minimize the expected evacuation time (von Schantz and Ehtamo, Physica A, 2022). The framework can be applied to passenger terminals, too. There, the crowd traffic is highly fluctuating, crowd size is uncertain, and various dangers like bomb threats, although rare, are of concern (von Schantz et al., Collective Dynamics, 2021).

Journal papers
A. von Schantz and H. Ehtamo (2022) Minimizing the evacuation time of a crowd from a complex building using rescue guides. Physica A: Statistical Mechanics and its Applications 594, 127011.
A. von Schantz, H. Ehtamo and S. Hostikka (2021) Minimization of mean-CVaR evacuation time of a crowd using rescue guides: a scenario-based approach. Collective Dynamics 6, pp. 1-29.
A. von Schantz and H. Ehtamo (2019) Pushing and overtaking others in a spatial game of exit congestion. Physica A: Statistical Mechanics and its Applications 527, 121151.
A. von Schantz and H. Ehtamo (2015) Spatial game in cellular automaton evacuation model. Physical Review E 92, 052805.
S. Heliövaara, H. Ehtamo, D. Helbing and T. Korhonen (2013) Patient and impatient pedestrians in a spatial game for egress congestion. Physical Review E 87, 012802, pp. 1-10.
S. Heliövaara, J-M. Kuusinen, T. Rinne, T. Korhonen and H. Ehtamo (2012) Pedestrian behavior and exit selection in evacuation of a corridor - An experimental study. Safety Science 50, pp. 221-227.
S. Heliövaara, T. Korhonen, S. Hostikka and H. Ehtamo (2012) Counterflow model for agent-based simulation of crowd dynamics. Building and Environment Vol. 48, pp. 89-100.
H. Ehtamo, S. Heliövaara, T. Korhonen and S. Hostikka (2010) Game theoretic best-response dynamics for evacuees' exit selection. Advances in Complex Systems Vol. 13, No. 1, pp. 113-134.
T. Korhonen and S. Hostikka (2009) Fire dynamics simulator with evacuation: FDS+Evac: Technical reference and user's guide. Technical Report VTT.
D. Helbing, I. Farkas and T. Vicsek (2000) Simulating dynamical features of escape panic. Nature 407, pp. 487-490.

Conference papers
A. von Schantz, H. Ehtamo and I. Pärnänen (2017) Twotype multiagent game for egress congestion. Hawaii International Conference on System Sciences, HICSS-50.
A. von Schantz and H. Ehtamo (2014) Cellular automaton evacuation model coupled with a spatial game. Lecture Notes in Computer Science Vol. 8610, pp. 371-382.
T. Korhonen and S. Heliövaara (2011) FDS+Evac: Herding behavior and exit selection. Fire Safety Science Proceedings of the Tenth International Symposium pp. 723-732.
H. Ehtamo, S. Heliövaara, S. Hostikka and T. Korhonen (2010) Modeling evacuees' exit selection with best response dynamics. Proceedings of Pedestrian and Evacuation Dynamics 2008 Wuppertal, Germany. Springer, pp. 309-319.
T. Korhonen, S. Hostikka, S. Heliövaara and H. Ehtamo (2010) FDS+Evac: An Agent Based Fire Evacuation Model. Proceedings of Pedestrian and Evacuation Dynamics 2008, Wuppertal, Germany. Springer, pp. 109-120.
T. Korhonen, S. Hostikka, S. Heliövaara, H. Ehtamo (2008) FDS+Evac: Modelling Social Interactions in Fire Evacuation. Proceedings of 7th International conference on Performance-Based Codes and Fire Safety Design Models, Auckland, New Zeland, pp. 241-250.
T. Korhonen, S.Hostikka, S. Heliövaara and H.Ehtamo (2007) Integration of an Agent Based Evacuation Simulation and the State-of-the-Art Fire Simulation. Proceedings of the 7th Asia-Oceania Symposium on Fire Science & Technology, Hong Kong.
T. Korhonen, S.Hostikka, S. Heliövaara, H.Ehtamo and K. Matikainen (2007) FDS+Evac: Evacuation Module for Fire Dynamics Simulator. Proceedings of the Interflam2007: 11th International Conference on Fire Science and Engineering. London, UK, pp. 1443-1448.

Doctoral theses
A. von Schantz (2021) Numerical simulation and optimization models for socio-dynamical features of crowd evacuation, Ph.D. thesis, Aalto University, Finland.
J.-M. Kuusinen (2015) People Flow in Buildings - Evacuation Experiments and Modelling of Elevator Passenger Traffic, Ph.D. thesis, Aalto University, Finland.
S. Heliövaara (2014) Pedestrian Behavior in Evacuations - Simulation Models and Experiments, Ph.D. thesis, Aalto University, Finland.
Open source project

In addition to the fire and crowd dynamics simulator FDS+Evac, we have also developed an open source Python simulation environment for studying crowd dynamics. Familiarize with it and our other research codes on our GitHub research organization page.