Learning Galapagos

Hi everybody,

I'm starting now to use Galapagos and looking around the blog and google I found some interessant exercises to follow. But also if they are easy I'm getting some problems or maybe is better to call them bug. Yes because when I run Galapagos to solve the problem it sometimes solve the problem completely but other times just in part. The problem is that with this kind of exercises that I'm doing you can check that there is a problem, but I'm studying it for my thesis and the goal will be to optimize a building in a ecological way, and the inputs will be more and sure will be hard to understand if Galapagos is working in the right way or not.

Therefore, what I want to ask is if someone can maybe check my files that I will upload and tell me if it run in the right way, maybe also trying it different times because like I said sometimes works in the right way and others not, leaving an element outside the calculation.

The simulation to run in the files are two, in the first there are 4 squares which I calculate the sommatory of the areas using the REGION UNION and doing it Galapagos should optimize the value giving me the minimum area between the squares. Sometimes it calculate it leaving apart one square. Anyway the right results should be to overlap all the squares.

The second one I have a catenary curve and 3 points and moving some slides Galapagos should adapt the curve to this 3 points. This simulation didn't give me any problem, so the problem should be the first.

Here is the definition of both the simulation.

I hope someone can solve my problem.

I also hope this exercises could be usefull for someone that is also starting to use Galapagos!

Thanks for your time and I will wait for your answers.

Attachments:

Senzatitolo.3dm, 152 KB
Trying with galapagos.gh, 15 KB

Replies to This Discussion

Permalink Reply by David Rutten on June 18, 2016 at 10:19am

Sometimes it calculate it leaving apart one square. Anyway the right results should be to overlap all the squares.

That is a side effect of the nature of your fitness criterion. You see if multiple squares are far apart, then moving them slightly will not affect the fitness at all. It is only when squares are already intersecting that a change in genome actually makes a difference. Galapagos will try to minimise the fitness, but it doesn't know what the best possible fitness could possibly be so after a period of ineffectually moving one of the squares about and not seeing any improvement it thinks it's done.

Basically this is the Galapagos equivalent of an underconstrained problem. One thing you can do is add an additional fitness element which pulls squares together, for example you could minimise the distances between all square centres. Even if this secondary effect is very weak, it will still pull the remaining square towards the others.

Another approach would be to restate the problem not as the minimum of the region union of the shapes, but as a minimising area of the Convex Hull containing all shape vertices. It might even be faster to compute.

Permalink Reply by David Rutten on June 18, 2016 at 10:50am

Here's another approach to the squares problem.

Attachments: