Metro Station Vulnerability Analysis with Crowd Simulation

Metro Station Vulnerability Analysis with Crowd Simulation


STAM is a multidisciplinary engineering company based in Italy that provides high-tech custom solutions to meet its client’s business challenges. The company leverages its expert knowledge from across engineering domains, including the security and transport sectors, sustainability and the circular economy, defense robotics, industry 4.0, and more. The company is an active participant in European Union research projects.


Due to terrorist attacks in Europe in recent years, the European Commission developed its funding opportunities for projects that improve public security and counter terrorist activities.

STAM worked on one such project. Its engineers created a simulation model of metro station crowd behavior for the purpose of analyzing different attack scenarios.

The objective of the crowd modeling was to simulate passenger behavior in metro station infrastructure during its normal state and facilitate comparison with various attack scenarios. Engineers also wanted to reveal and analyze any critical vulnerabilities in public infrastructure, using the capabilities of the model for evacuation simulation.

Modeling people's behavior in a standard conditions
Modeling people's behavior in a standard conditions (click to enlarge)


Engineers simulated an open-air metro station with two independent entrances and two floors. The train platform is located on the upper floor. The model also included station facilities that can be used by pedestrians during their travel time, such as ATMs, toilets, a coffee point, and shops.

Within the framework of the model, STAM engineers created three metro station attack scenarios:

Each type of attack was represented in the model as an agent. For each agent, engineers created a statechart that represented the different phases of an attack. And, by combining statecharts with flowcharts, the agents could move around inside the simulation environment. The only agent without a flow chart was the explosive agent, due to it not needing to move and being represented as a static agent.

Statecharts and flowcharts for the agents
Statecharts and flowcharts for the agents (click to enlarge)

Using AnyLogic’s 2D and 3D animation capabilities makes it possible to observe simulation model runs in many different ways, allowing analysts to examine agent behavior for each scenario in a simulation environment.

Simulation of the scenarios
Simulation of the scenarios

For the first scenario, with the explosive, the model showed the effects of the bomb’s explosion and the response triggered by it, including the reaction times of pedestrians in the vicinity and their evacuation.

For the scenario with the melee weapon, the model revealed that crowds took longer to react and evacuate due to the localized character of the attack. Also, the model showed the movement of the terrorist trying to leave the station and being stopped by security.

Finally, in the third scenario, with the drone, the attack happens while a train is arriving at the station – when the amount of people on the platform is at a maximum.


The density maps associated with the crowd modeling helped analysts identify critical areas for pedestrian flow and the points where potentially dangerous situations may arise.

All scenarios provided results for comparison, showing the number of casualties, injured, and exposed, as well as the overall evacuation time and critical areas.

Simulation results
Simulation results (click to enlarge)

The results of the crowd modeling show that the average time of evacuation is approximately the same regardless of the scenario. The explanation is simply that the metro station infrastructure is the same for all scenarios. The capacity and evacuation routes are the same in each case.

Although evacuation times are similar, the number of casualties is not and depends on the type of attack. The drone attack caused the greatest number of casualties and injuries due to its targeting of a train arrival, when pedestrian numbers are at maximum density.

The company continues to collaborate with EU Commission on the terrorism safety project.

Furthermore, STAM plans make additions to their project. For example, a smart signaling system to improve evacuation times. They also intend to implement and simulate security systems that are based on artificial intelligence algorithms that can recognize suspicious objects, such as a bomb. Evaluating these systems can determine how they can best help prevent an attack.

The case study was presented by Pietro De Vito, of STAM, at the AnyLogic Conference 2021.

The slides are available as a PDF >>

Similar case studies

More case studies

Get a brochure with industry case studies