How much upwind progress does a sailing drone lose to tacking?

A wing-sail drone polar and three tack controllers on a month of NOAA wind: a boat that watches only the wind ahead loses about 12% of upwind progress.

A small autonomous sailing boat that must reach a mark upwind has to tack, and the wind does not hold still. How much progress does it give up, and does tacking when the wind heads the boat help?

I built a first-order polar for a 5 m wing-sail drone: a finite-wing lift model with stall, balanced against a Wigley hull (the Michell wave resistance already in research/marine-hull) and a keel. Its inputs are swept, not taken from a vendor polar. Over 200 random drones the best upwind angle at 5 m/s has a median of 40 degrees (range 30 to 52; 20 of 200 fell outside the 35 to 55 I predicted). With hull drag set near zero the model's apparent wind angle equals atan(D/L), as it should.

Then I ran three controllers on 24 hour windows of NOAA CO-OPS 6-minute wind, 12 station-months in January and 12 in July 2024, holding a 1.5 km corridor about a mark straight upwind, with a 30 s loss per tack. In the July holdout, as a share of the best progress possible with no corridor and no tack cost:

  • tacking at the corridor edge: 83.7%
  • tacking when headed (margin picked on January): 87.6%, +3.2 points, better at all 12 stations
  • a clairvoyant that knows the whole day: 99.7%

I predicted the first would reach 85% and the clairvoyant would be within 5 points of the header-follower. Both were wrong; the gap is 12 points. The clairvoyant also exploits gust noise in the samples, so 12 points overstates what a forecast could give. The wind here is uniform in space and the stations are harbours, so this is a study of shifts over time, not weather routing.

Code, data and the full log: research/marine-hull (LOG.md, 2026-11-04, milestone 5a). Simulation only; nothing was tested on water.

All writing · More in Research findings · More on marine-hull · All tools and research