Integrating Bayesian Networks and Pathfinder Simulation for Risk-Informed Evacuation Planning
Author
Folden, Tobias Rosenørn
Term
4. term
Education
Publication year
2026
Abstract
This thesis addresses evacuation challenges in historical, geometrically constrained venues where traditional design changes are limited. It proposes an integrated framework that combines a Bayesian Network with the Pathfinder evacuation simulator to link uncertain human behavior with spatial egress dynamics. The Bayesian Network quantifies stochastic behavioral parameters—specifically walking speed and initial delay—which are used as inputs for agents in Pathfinder. Tivoli Gardens’ layout was mapped using satellite imagery to create realistic scenarios, and Monte Carlo simulations within Pathfinder enable a robust assessment of how behavioral variability affects overall evacuation times. In a Tivoli Fredagsrock case study, results indicate that safe evacuation (under 1,200 seconds) becomes highly improbable when attendance exceeds 10,000 people. The framework was also used to test non-spatial mitigation measures and shows value for risk-informed crowd safety planning, supported by a problem analysis including system decomposition, stakeholder mapping, and pertinent evacuation theory.
Dette speciale adresserer udfordringen ved evakuering af historiske og geometrisk begrænsede venues, hvor traditionelle forbedringer af udformningen er vanskelige. Studiet foreslår en ramme, der integrerer en Bayesian Network-model med evakueringssimulatoren Pathfinder for at koble usikker menneskelig adfærd til rumlig egress. Bayesian Networket kvantificerer stokastiske adfærdsparametre – især ganghastighed og initial forsinkelse – som anvendes som input til agenter i Pathfinder. Tivoli Gardens’ fysiske layout blev kortlagt med satellitbilleder for at sikre realistiske scenarier, og Monte Carlo-simuleringer i Pathfinder muliggør en robust vurdering af, hvordan variationer i adfærd påvirker samlede evakueringstider. I en case om Fredagsrock i Tivoli viser resultaterne, at sikker evakuering (under 1.200 sekunder) bliver meget usandsynlig, når publikum overstiger 10.000 personer. Rammeværket blev desuden brugt til at afprøve ikke-rumlige afværgeforanstaltninger og fremstår som et nyttigt værktøj til risikobaseret planlægning af publikumsikkerhed, understøttet af en problemanalyse med systemopdeling, interessentanalyse og relevant litteratur og teori om evakueringskriterier.
[This apstract has been generated with the help of AI directly from the project full text]