Sizing a cargo-drone hub's spare batteries with a formula instead of a simulator

For a battery-swap drone hub with 8 chargers, a short queueing recursion gives the same spare-pack count as the simulator: 12 to 45 packs.

The aerolog-scaling programme has been sizing a battery-swap hub for a cargo VTOL by simulation: 8 pads, 8 chargers, a 42-minute charge and 3 minutes on the pad. Each answer took a grid search over pack counts. This milestone asks whether a formula gives the same answers.

The rule: a flight waits for a pack when the number of packs not yet charged reaches the pool size S. Packs at the chargers form an M/D/c queue (constant charge time). Its distribution follows exactly from a one-step recursion, N(t+D) = max(N(t) − c, 0) + arrivals in D, the method Crommelin used in 1932. Packs between the pad and the charger add a Poisson count. In pure Python the whole calculation takes half a second.

Smallest pool with at most 5% of orders waiting:

Hub loadOrdersFormulaSimulator
70%single1212
70%pairs1616
70%batches of 42525
85%single1717
85%pairs2626
85%batches of 44545

Two of my three predictions failed. I expected the formula to err on the safe side, and it errs on the other. A flight that waits sends its empty pack to the charger late, which makes the next shortage worse. With the pool within about 2 packs of the mean count, the formula under-states waiting by up to 40 points; pools that small are far below anything you would build. Second surprise: batches of 8 orders at 8 pads leave about 5% of orders waiting however many packs the hub holds. That is the chance that a second batch lands while all the pads are busy, a pad limit and not a battery one.

Limits: constant charge time, fixed batch sizes, steady demand. Code, results and log: research/aerolog-scaling/ (milestone 6i).

All writing · More in Research findings · More on aerolog-scaling · All tools and research