What made the puzzle unusually difficult was the fact that the pieces were not in the rows and columns most puzzle use. Instead they seemed to be based on several large arcs the were randomly placed within the puzzle boundaries. Also, the complex shape of each piece added to the difficulty.
I Googled “how are jigsaw puzzle pieces made” and the answer was they are “designed using a CAD program…and cut by a machine with very sharp edges.” Well OK, which program might that be? And how would one go about tiling a rectangular area with 1000 pieces that don’t seem to be randomly situated and that look like those in the above image?
My first thought was Kangaroo, but I don’t know enough about how it works to even know where to start. Obviously there is some randomness involved, but I don’t see how it might be applied. Maybe someone knows the secret behind how the shapes are made and can explain the method? I’d love to be able to 3D print shapes like this - is it possible?
Dear Birk, from the forum I know you have more ideas then “what CAD could it be?” and maybe “Kangaroo” Nice challenge though
In the real world, what would the “cutter” look like. Maybe long bend “razor” steel, hold in a 3D printed plate. All corners are sharp (in your picture) so where the lines cross, they were soldered/welded and sharpened? I guess a lot of manual work, to get this thing cutting nicely.
Or maybe “laser” cutting? That would be more accurate. Maybe someone can jump in and tell us: “How this is made”.
For the design of the cutting pattern, why not start with some designing rules:
The tabs and slots are not al the same size, maybe they are all different? Also the lines, some are longer (like they were designed like that) other short. Anyway there is this combination (balance?) between order and randomness.
The tabs being irregular, could be the effect of bending steel strips. If designed, they will have max/min sizes, so the pieces will not break, but fit.
there seems to be a pattern in slots/tabs slot - tab - slot -tab. But not always. I also see 3 tabs in a row.
every piece has 4 or 5 tabs, and so every piece can be seen as an irregular quadrilateral or pentagon.
I think, discussing this a bit, that more and better rules will come up.
Why not in grasshopper?
Bottom up (from small to big): Point cloud → randomly 4- and 5 sided polygons → grow them together? In between curves → resizing, untill all is filled.
Then explode in sections, middle curve, set tab/slots. Within the rules: 4 sided = 2 tabs+ 2slots: 5 sided 3 +2.
etcetera
Or top-down:
pattern of curves, there must be randomly methods to create them? then shorter curves to connect curves, untill we have small surfaces with minimum/maximum sizes (puzzle-piece).
Creating tabs and slots again.
But surely, you already came up with all of this, and probably better,
Regards, Eef
Eef Weenink: I just knew there would be people here who could point to or give solutions for my questions, even though I had never seen anything about this matter before. And as usual the responses were far better than I expected.
I hoped I was on the right track (GH + Kangaroo) but my inexperience with Kangaroo left me with no good idea about where to start. I had thought about Voronoi methods too, but now I’ve got several good starting points. It’s going to be interesting to see what I can come up with.
**Laurent Delrieu: what can I say - your work is always spectacular and provides great pointers on how to proceed. The Nervous site is quite impressive for sure, and I was surprised they let their cat out of the bag by publishing this: “**The pieces are grown in a simulation of elastic rods. The edges grow, lengthening, until they collide, pushing each other into contorted shapes.” That’s a perfect recipe for Kangaroo of course - so now I just have to cobble up something that works.
Their blog/Custom Puzzle section shows how Kangaroo could grow a Voronoi pattern into nicely shaped puzzle pieces, so I was happy my original thoughts were on the right track.
Martin Siegrist: that’s a great find showing how a semi-manual method could be done. Of course only the Chinese would think about doing it that way, but they sure did nail the overall technique. It makes sense they would follow the row/column approach considering how they manipulate the cutters - again a very Chinesey approach using their metal bending device - but for a 3D printed solution I plan on focusing on the Kangaroo method of manipulating rods (filleted rectangles) or Voronoi patterns.
Thanks again for all this insight. I should be pretty busy for a while now.
I did a puzzle series a few years back. trick was finding ‘nodes’ in between curves, then dividing the curves between the nodes into a consistent unit length. As I recall I laid out the curves by hand; once I had the overall curve I could then superimpose the keyway curve.
Kangaroo to collapse points, turning n-gons to 3,4 or 5-gons. Doesn’t work very reliably at all. Some pieces get too thin, and some regions fail completely, despite my best effort to purge any 0-length lines from the Kangaroo output.
Assuming that can be fixed, making the tabs and holes between aligned pieces should be simple (explode all boundaries, purge duplicates, orient and scale random tabbed curves to replace each line), but the offset joints along the major “fault lines” would be very challenging, to make sure no tabs are too close to the corners. It’s still not quite like the example image, which has T-junctions within each region as well as along the region boundaries. Maybe kangaroo should also be applying some forces to try and make the lines at each junction come together at 90° intervals or something?
Nice challenge though 
