A friend of mine is sending an unmanned autonomous boat over the Atlantic ocean (https://projectmahi.com/).
I’ve been building a keel prototype for them.
https://twitter.com/finfoil/status/977803040311922689
Designed with finfoil even the CNC toolpaths are generated using experimental finfoil code.
That looks really nice Hans. The code for the CNC was looking very nice the last time we talked.
Did you do anything special with the plywood before you milled it?
Sometimes I have laminated a couple layers of glass cloth and resin to the down side (inside of fin when completed) to keep the wood from splintering on the tip and trailing edge during milling. But often I am not patient and I just make the parts.
Is that the final attachment method or just to accommodate windsurfing for testing 
?
I did not do anything special to the plywood, it was standard 17mm plywood, nothing special added.
Because I did not use tabs to hold the fin while cutting the outline, I told my gcode-generator my stock was 16.5mm, leaving 0.5mm attachment that I cut out using the oscillating saw you see in the second pic.
I have now implemented 3 different contour based strategies and have added a post processing step to offset the toolpath by the tool radius.
Currently implemented strategies:
-simple: just calculate a fixed amount of contours at equidistant thicknesses
-within: ensure overlapping contours using the boost geometry buffer and within functions (http://www.boost.org/doc/libs/1_66_0/libs/geometry/doc/html/geometry/reference/algorithms/within/within_2.html)
-distance: ensure overlapping contours using the boost geometry distance function (http://www.boost.org/doc/libs/1_66_0/libs/geometry/doc/html/geometry/reference/algorithms/distance/distance_3_with_strategy.html)
Both the distance and within function seem to fail to ensure perfect overlap (likely to be caused by some low resolution contour calculation used in the solver to make it fast :P).
So I’m now working on a top view visualization of the material removal so I can figure out how to tune the algorithms for speed and accuracy.
All this will be made available at some point, but you can always contact me for the generation of specific toolpaths.
Hi Hans, thanks for the detailed explanation. The programming and math goes way above my end-user head.
I am working on one at work. I am on my third or fourth try and have gotten the time down from 6 hours using simple software to 3 hours using code generated by 2 different programs: one to dynamically rough out the part and another to do a very clean line-arc-line finishing pass. For this situation I could no longer rely on STL and low-cost CAM as the faceting and rounding was producing flaws.
A good algorithm would also be important for those of us using the small cutters on home machines.
Thanks again for sharing your progress and offering it to us.