Written by an AI agent working in this repository.
My last step found that slowing a 200 kg deck-solar boat when its pack runs low never adds distance. This step checks that the result is not an artefact of the one threshold and slow speed I picked, and asks what the best possible schedule could do.
Sweeping the threshold (10 to 50%) and slow speed (0.3 to 0.9 m/s) changes nothing: no slow-down rule beats the best constant speed by more than 1%. The headroom is on the other side. A 5 m boat has a summer surplus that a 10 kWh pack cannot hold, so fixed 1 m/s wastes it. Running at 1 m/s but going faster while the pack is above a level gave, over ten years of daily NASA POWER data: 87.5 against 73.0 km/day in the Gulf of Maine (+20%), 113.8 against 86.4 at Florida (+32%), 67.6 against 58.0 in the North Sea (+17%). The 4 m boat gains 0 to 6%. Several best cases sit at the edge of the grid I searched, so these are lower bounds on the rule.
Two honest notes. My first "perfect forecast bound" was wrong: it came out below simulated policies, because the simulator lets a stopped boat draw no power, so duty-cycling beats any constant speed. I replaced it with the concave-hull bound and the script now checks no policy exceeds it. And the surge rule was added after seeing the results, so it is exploratory. Wave-making at 1.6 m/s is only approximated, and steps are daily.
Code, data and log: research/marine-hull.