generative modeling for Rhino
I was wondering how I can improve the performance of Galapagos in an exemplary setting.
My exercise is composed of stacked programmatic units (housing and office) referenced as breps and interpreted with horster Reference. Furthermore, a triangulated facade covering the units is evaluated regarding solar access performance with ecotect via geco. Since the energetic demand of each facade panel is determined by the adjacent program unit, a fitness ratio can be easily calculated (demand - calculated solution). Finally, I'm adding up all ratios and feed them into the solver as value to be minimized.
However, I'm not really satisfied by the 'optimized' solution. The solver record and optimized geometry are still object to improvement. See files attached.
Thanks in advance,
You have a lot of parameters in the gene that can vary. Less genes mostly lead to faster convergence with the evolutionary solver. Then your genes are likely to produce a similar value of fitness with totally diffenent configurations. Galapagos will have a hard time to settle on one solution with the evolutionary solver.
Instead of varying all the panels separately, I'd rather use 3 or 4 attractors and move them around. This might give a little less random results.
here is something about multiple sliders used with genetic algorythms...