02 / PEOPLE OUTSIDE THE BUILDING
03 / FINAL ROUTE COUNTS
Monte Carlo results · 300 runs per situation
These averages describe repeated simulations. The animation above shows a deterministic reference run, so its completion time differs.
Guidance comparisons & cost
Paired differences relative to the corresponding concert or lecture baseline. Negative values mean a shorter time.
Cost / time combinations retained in the archived cost analysis.
Model, movement and wheelchair assistance
The final Python model combines a building network, route choices, integer population transfers and capacity-limited queues. Normal dismissal uses truncated Gaussian reaction times, concert emergencies use lognormal reaction times, and lecture emergencies use Gamma reaction times. The playback contains the nine saved situations from the final experiments.
The seating rows, aisles and stair drawings retain the original visualization style. They are schematic. Each displayed person's transfers follow the model's recorded whole-person flows; paths between display positions are illustrative and do not constitute a separate pedestrian simulation. Display identities are assigned in FIFO order and do not identify the original reaction-time or mobility cohorts.
Three wheelchair users are included in M2, replacing three walking occupants. The reconstructed dots do not distinguish their identities. The model's wheelchair refuge and assisted street-level times are shown for each situation. The front exterior steps do not add delay for walking occupants; side exterior stairs remain in the model.
Concert scenarios use six Crowd Managers at the model's baseline locations. LED and barrier cases use their saved experiment settings. The blocked-exit case closes all five main doors, with a 20 s information delay. Building evacuation is measured until the last person leaves the building; it is separate from reaching street level.
Results use assumed passage dimensions and behavior parameters. The 180 s line is the problem's comparison target. See the public repository for the model and saved experiment results.