That's equivalent to calculating the "straight skeleton", which is mentioned in the article as being among the published solutions, but too slow and error-prone

That's equivalent to calculating the "straight skeleton", which is mentioned in the article as being among the published solutions, but too slow and error-prone. https://en.wikipedia.org/wiki/Straight_skeleton https://en.wikipedia.org/wiki/Straight_skeleton There's no explicitly spiraling/lawnmower aspect to the straight skeleton: you may have a raster-based incremental literally spiralling solution in mind, but for the geometric examples they show this would probably be inefficient. It's going to have to traverse a large multiple of the perimeter of the polygon as it spirals.