©icd/itke, university of stuttgart

©icd/itke, university of stuttgart

©icd/itke, university of stuttgart

Project N°002

Name Textile Tango

Program Master’s Thesis Project

Year 10-24

Status Built

About

This research investigates how occupants in a building respond to a spreading fire by
considering their traits and surrounding conditions in a fire scenario simulation. The study specifically assesses the impact of navigation sign placement on fatality and survival rates.

In order to model basic assumptions and behaviours, we referred to case studies on the
subject; this is particularly helpful since the evacuation problem we are addressing is a
common and well-established application for agent-based systems.

To construct a robust model, we incorporated insights from existing studies on evacuation
problems, particularly those utilizing agent-based systems. Leveraging this knowledge, we
developed a simulation framework featuring two interconnected agent systems: "boid" agents representing occupants and "meshFace" agents serving as the environment.

Within our simulated environment, initially represented by a typical floor plan containing walls,
obstacles, and visual indications of escape routes, the meshFace agents act as a vital
component, influencing the attributes and behaviours of the more complex boid agents. The
meshFace agents play a critical role by providing fire values that affect the health status of the
boid agents.

The boid agents, characterized by age, gender, and health status, exhibit intricate behaviours
such as cohesion, alignment, separation, and response to fire levels. Their evacuation
behaviours are triggered by an "isAlarmed" boolean, determined by the proximity of meshFace
agents exceeding a predefined value. This influences the health and speed of the agents,
ultimately affecting their chances of survival.
Our simulation introduces a level of realism by considering the agents' field of view, depth, and
sight limitations imposed by obstacles. Through iterations and weighted distributions of
behaviours, we arrived at a model that produces plausible results in various scenarios,
acknowledging the potential for further fine-tuning.
The simulation begins by creating agents with specific characteristics, and their responses to
the environment are determined by their health status and proximity to fire. Evacuation
behaviours are activated when agents are alerted, leading them to follow navigation signs and
other agents towards exits.
To validate our model, we conducted simulations with randomized navigation sign placements
on walls, recording key metrics such as average health, and counts of alerted, dead, and safely
harboured occupants. We also explored the impact of different fire start positions on
evacuation outcomes.
Looking ahead, our research roadmap involves comparing results from different navigation
plans and assessing the robustness of these plans against varied fire start positions, possibly
benchmarked against codes and best practices. Furthermore, we envision connecting our
model to an optimization algorithm to enhance navigation sign placement strategies, potentially
improving overall evacuation effectiveness. This research lays the groundwork for developing
advanced navigation planning tools crucial for enhancing fire safety protocols in buildings.

©2024 Yağmur Bulut. All rights reserved.

//////////