You’re utilizing Generative Design to solve a puzzle (only one solution) and so it’s giving you only one solution. Feed it a problem (many possible solutions) and it should give you many outcomes. Feed it a wicked problem and you’ll find even more value in exploration and in running multiple studies.
The Interactive Design Foundation has a decent primer of wicked problems here: What Are Wicked Problems. But to put my spin on each of these:
- Puzzle: Something with only one way to put it together.
- Simple Example: sudoku; there is only solution for any given set of outcomes.
- Design Example 1: Placement of the fire control center (the jurisdiction I practiced the most in these HAD to be off the lobby and directly connected to the exterior).
- Design Example 2: Sculpting exterior panels to reduce solar heat gain.
- Problem: Something with many possible solutions, the ‘best’ of which is often subjective.
- Simple Example: the tanogram where there are (basically) limitless ways you can configure them, each no better or worse than the last as there is no ‘one shape’ you’re after.
- Design Example: Sizing of a structural beam; Many shapes will be sufficient, but one might be better than the others due to the profiles/depths already used on the job.
- Wicked Problem: A problem with no end or solution, where there are no constants, and you will be wrong at some point.
- Simple Example: Trying to solve a sudoku board where the starting values change every 5 seconds.
- Design Example: Design of a building, as the occupant needs aren’t static, and someone will be unhappy with the result.
I can’t open Dynamo right now as I’m on the other laptop, so I may have some incorrect assumptions here. However from what I can see you’ve presented the tool with a problem, and it’s given you the solution. "Find the shape which tessellates the surface in the way with the least amount of wasted panels while providing for variation between shapes.
Ask instead for it evaluate the panel shape which has the least amount of wasted material AND the most consistent panel shape.
- Wasted material would result in something like what you’ve presented above.
- Consistent panel shape would result in squares.
What is the mix of those two? Well there likely isn’t one solution for that.
Adding additional outcomes (fit to catalog size; minimum and maximum dimensions, structural capacity (acute angles are more apt to snap off in shipping/handling/installation) and suddenly you’ve got a situation where there isn’t a solution which makes everything happy. Now you get to work at weighting those evaluations either manually (make decisions primarily on the X axis, but account for the Y and the color and the size while having filtered out results which didn’t achieve another axis), or computationally by weighting your outcomes (multiply/divide values based on conditions being met or not).
Note that just because something is a puzzle and not a problem or wicked problem doesn’t mean that the use of a generative design process (in any platform) isn’t valid. Both of the ‘classic’ examples which @GavinCrump provided are technically puzzles - there is one orientation and location for a building which has the best environmental performance; there is one shape of panel which reduces solar heat gain the most. But the generative process (an in some cases only surfacing one solution thereof) is ideal there as manually exploring the design space would take a lifetime to execute. After all there are 360 possible rotations of the building (assuming you cap at 1 degree), and (assuming a 1% step in the U and V direction) 100 possible X values and Y values, resulting in 3,600,000 possible ways to place the building on site. That’s several days worth of studying and cataloging input > outcome relationships, vs a half day to write a tool which you can run while you work on another design problem.
