Positioning the fleet…
The warehouse locations are fixed and identical in both panels. The decision is how many of the six vehicles to station at each of the three depots before orders arrive. Both sides use the same road network, vehicles, arrival times, deadlines, and dispatch rule. The signs above each warehouse show its vehicle count. Each delivery requires a return to its assigned depot. Floating numbers show outstanding requests at each stop; a plus sign marks a recent arrival, and red means at least one request is overdue.
Minimizes expected delay under the nominal forecast: 85% clear days, 5% eastern storms, 10% western storms.
Minimizes worst-case expected total cost over every probability law within L1 distance 0.30 of the nominal — at most 0.15 of mass moved.
Minimizes worst-case expected excess delay over the best fixed split under the same law — the part of the loss a different split could have avoided.
The storm slows the bridges and shifts orders toward one region. Vehicle positions remain fixed; neither dispatcher sees future orders. The right-hand result is not guaranteed to win every weather scenario.
This is a constructed, deterministic illustration—not a calibrated logistics model or empirical validation. One playback second represents one operating minute. Each order has a 14-minute deadline. Cost accumulates one unit per minute of waiting plus four per minute overdue, including undelivered orders.
Built by Elioth Sanabria — more systems are coming. Follow @elioth4u for the next one.
A Staffing Analytics project. We build interactive decision tools like this for operations and analysis teams — contact@staffinganalytics.io.