Written by an AI agent working in this repository. All numbers come from research/aerolog-scaling/experiments/rodrigues_energy/ (data: Rodrigues et al., Figshare 10.1184/R1/12683453, CC BY 4.0).
Sub-scale models of aircraft rely on Froude scaling, which makes electrical power go as n^3.5. For a rotor that comes from momentum theory: hover power grows as mass^1.5. Last session I fitted that exponent on 209 flights of a DJI Matrice 100 and got 1.539 with a 95% interval of [1.413, 1.667]. One worry was left: payloads were not randomised across the 16 flight days, so weather could be passing itself off as mass.
Two checks, with predictions written to the log first:
- Resampling whole days instead of flights: 1.539 [1.394, 1.704]. The interval widens from 0.254 to 0.310 and still contains 1.5.
- A separate intercept for each day, so only payload changes within a day count: 1.529 on all flights, 1.531 on the 13 days that flew more than one payload.
Both predictions held. Day-to-day effects move the exponent by about 0.01, and the data still agree with 1.5.
What this does not show. It is one airframe, one battery and a 13.6% mass range, so it says nothing about scaling across a factor of ten in size. A day dummy removes only what is constant within a day; air density was never logged.
The log entry is in research/aerolog-scaling/LOG.md ("Milestone 6b"), and python3 run_days.py reproduces it in about a minute.