Multi-Scale Hospital Evacuation Simulation

Emergency Department, Johns Hopkins Hospital


Scenario & ED specs:

  • Time: 12:00 AM (Midnight)
  • Total number of patients: 80
  • ED units: Psychiatry, Main ER, EACU, RAP, Triage, Admissions, Trauma
  • Total number of nurses: 37 (with specific assignment to units)

Agent-Based Model:

  • Types of patients: mobility impaired (wheelchair users, high-acuity bed-bound, low-acuity bed-bound, and stamina impaired), non-disabled (adult and elderly)
  • Path planning: Floyd-Warshall Algorithm
  • Collision avoidance: The Predictive Collision Avoidance Model developed by Karamouzas et al. (recalibrated for evacuation behavior)
  • Social behaviors: information sharing, grouping, herding, and rescuing

Fluid Dynamics Model:

  • Introducing the modified Payne–Whitham fluid dynamics model (PW++) with multi-type densities and a meso-scale model of social behaviors
  • Dynamic potential field solved using the Fast Sweeping Algorithm for Eikonal Equations
  • Congestion avoidance implemented by considering density in the penalty function of the potential field

Stock-and-Flow Model:

  • An state-of-the-art SF model
  • From ABM to SFM: ABM reduced to an exit rate function using Monte Carlo simulation and nonlinear regression, estimating exit rate based on aggregated mobility characteristics of evcuees and overall characteristics of the room

Reference Papers

ABM