Emergency Evacuation Simulation
A large-scale agent-based simulation modelling emergency evacuation dynamics across five university lecture-hall layouts. The system incorporates heterogeneous agent populations — students, professors, mobility-impaired individuals — each with distinct panic thresholds, familiarity coefficients, and movement speeds. Core engine: a Dijkstra-variant pathfinding algorithm with real-time rerouting under congestion. Panic contagion and herding dynamics are modelled as emergent social phenomena. Over 9,000 parameterised experiments conducted via BehaviorSpace.
Key finding: Modern multi-aisle layouts with symmetrical exit placement reduce median evacuation time by up to 38% versus traditional single-aisle configurations. Herding behaviour degrades performance even in well-optimised layouts.
- NetLogo
- Python
- Pandas
- Matplotlib
- BehaviorSpace
- Dijkstra pathfinding