Fixed center, measured in room units north of the door.
02 / MY PROJECTS
My Projects
Interactive models for questions that are easier to understand when you can see them.
OCTOBER 2026 · AI-ASSISTED MODELING
Room Evacuation &
Doorway Simulation
Can a fixed person near a doorway make a crowd leave faster? I collaborated with ChatGPT to build a simulation, refine its movement assumptions, and inspect how obstacle placement changes the outcome.
The room contains 50 moving circles. A fixed circular obstacle changes their paths; the study compares matched initial layouts with and without that obstacle.
WHAT THE CURRENT MODEL SHOWS
With the fixed person / without it, across 32 fresh layouts.
Mean time saved at this tested position, about 8.5% of the control mean.
These are recorded results from the corrected study embedded below. The selected position is the best among finite tested candidates, not a proven continuous optimum. Lateral placements can stall and very close placements can block the route.
Inspect the heatmap, select a tested placement, replay matched crowds, or run a fresh seeded pair. The full study contains the methods and numerical checks.
Explore the evacuation study
The original HTML includes heatmaps and detailed replay data.
Assumptions matter.
The square room is 2 × 2 units with a centered ¼-unit doorway. Each person has diameter ⅛ unit and a preferred free speed of ⅛ unit per second.
Obstacle avoidance uses tangent–arc–tangent paths for moving centers. Contacts are nonbouncy; steering, walls, and the immovable obstacle exchange momentum with the crowd. Actual motion can differ from the reference path.
This is an idealized driven-circle model, not a validated prediction of real emergency evacuation. The study checks symmetry and sensitivity to time step and routing rules.
← Back to all sectionsGUANHUA (LORENZO) DU