Emergency evacuation is a critical component of emergency response and requires developing in advance evacuation preparation activities ensuring people can get to safety in case of emergency. In order to define effective evacuation protocols, understanding disasters and crowd emergency evacuation behaviour are essential.
A full-scale evacuation demonstration is not viable because of ethical, practical and financial issues. Therefore, models and simulations of crowd behaviour are widely used to analyse the effectiveness of evacuation preparation activities. Multi-Agent Systems (MAS) have been proposed for simulating individual cognitive processes and behaviour, as well as for the exploration of social or collective behaviours. At a microscopic level, human individuals are represented as autonomous agents equipped with sensors, decision-making rules, and actuators. At a macroscopic level, human social behaviours are modelled through simulating the interactions among agents or groups in a virtual environment.
In the scope of CitiSim project, two specific simulation models related to the Smart Emergency use case have been modelled and implemented: the model of normal behaviour and activity of crowds in buildings, and the model of crowd behaviour in an emergency situation (including the modelling of the occupation decisions, the threats and the evacuation policies). Both models have been considered due to their importance in emergency situations in buildings.
The developed simulation software has been called SOBA (Simulation Occupancy Based on Agents) and is available with an open-source license on the Github platform. The goal of SOBA is to provide a simulation tool for modelling crowds in buildings. However, the study of situations related to emergencies in buildings requires an extension of this software. This extension has been called SEBA (Simulation of Evacuations Based on SOBA) and is a useful simulation tool for studies related to emergencies and evacuations in buildings.
In addition, we have considered the implementation of a specific visualization tool. This tool enables the visualization of social simulation in a 3D environment, and also the creation of the floor plan of a building. The tool has been called RAMEN and is provided as open-source on the Github platform. The integration of both tools, RAMEN and SEBA, enables an efficient and representative real-time and offline evaluation of simulations of crowds in buildings.
The system has been validated in a real-life scenario, the second floor of Building B of the School of Telecommunication Engineers (ETSIT) of the Universidad Politécnica de Madrid (UPM). This floor has a rectangular shape and an area of 1600m2 (~100m x 16m). In addition, in the building, there are four types of rooms: offices, laboratories, classes and transition spaces (such as halls, corridors or resting areas). Therefore, the floor is divided into 41 rooms distributed as 14 offices, 20 laboratories, 54 classes, 1 hall and 2 corridors. There are two available exits, one to exit from the building and another to enter to other connected building. In addition, the occupation is medium in offices (1-5 professors) and laboratories (3-8 researchers); low in transition spaces; and high in classes (30 students). An example of a simulation made by the platform formed by the simulation software and the visualization tool is available on the GSI website. In the video, it can be seen how visual representations of relevant elements have been implemented, such as the occupants, walls,
Figure 1: Example of evacuation simulation in the complete scenario
Figure 2: Example of evacuation simulation in the testing scenario
The building has been modelled in the simulation considering a space formed by rooms. In addition, the rooms are divided into positions of 0.5m x 0.5m, which is an approximation of the occupied space by one person. Each room is characterized by width, height, type (office, corridor, laboratory, class, hall or restroom), inner and external walls, and the connections between them, where doors are positioned.
Author: Sergio Muñoz López, UPM