Good evening, I have several questions concerning multi-goal-optimization. I´m trying to optimize a (kind of) complex problem with galapagos/wallacei/goat/etc. The definition attached is a very simplified version of my optimization problem that is sufficient to explain the problem:
Goal 1: Maximizing the area of a polygon (named A) that is contained in another (bigger) polygon (named B)
Goal 2: Minimizing the overlap/collision of a polygon (that is an offset of polygon A) with polygon B
In other words: The area of the inner polygon (A) has to be maximized while staying away from the outer polygon B (by a certain distance).
A solution very close to the optimum would look like this:
Obviously, the optimum can very easily be calculated by simply offsetting the outter polygon B by a certain distance to get the maximized (inner) polygon A. Like I`ve said my actual problem is more complex where the offset-distance will vary from point to point. So for now I want to stick with the solver-approach. My questions:
- Galapagos: Even though my optimization problem is very simple it takes galapagos 5+ minutes to get near the optimum not even reaching it. It seems to stick with a local optimum. What´s wrong with my definition? Galapagos actual just needs to “push” the inner polygon outwards until the offset “touches” the outer polygon?! Is my fitness-function set up correctly? That´s the solution galapagos sticks with:
-
I
ve realised that galapagos needs way less time if I "preset" the genome-sliders in a way that is somewhere close to the optimum. This way galapagos even finds the optimum. So instead of sending galapagos "in-to-the-blue" I can give it a hint where to go. Like a "initial population". Im just wondering why galapagos depends on this for such a simple task. -
Wallacei doesn´t seem to take the initial genome-state into account. Can I give wallacei a hint where to start optimizing from?!
Thanks in advance!
wallacei_04.gh (14.3 KB)