Some Galapagos & DIVA simulation questions

Hi. I'm using Galapagos and DIVA for my architecture thesis project. And I hope I could get some help.

In my project, I am trying to use Galapagos to maximize the amount of daylight in a high-rise building floor plan. This is done by minimizing the number of areas with less than a certain amount of lux. To simulate the daylight level or illuminance of the floor plan, DIVA was used.

Below is what my grasshopper definition looks like:

1. In Galapagos, what is the difference between the Evolutionary Solver and Simulated Annealing Solver?

(a) When I used the Evolutionary Solver, somehow there wasn't any results showing although I let the simulation run for quite some time. I could see that the solver was running as the DIVA simulation pop-ups would appear but in Galapagos, everything remained blank. Could anyone tell me why this happens? This is what it looks like:

(b) When I used the Simulated Annealing Solver, there would be results appearing for awhile and then suddenly there would be an error and the whole process would stop. The graph would disappear and the results visible but there are some problems retrieving the information. See below:

2. Also, is there any way I could save the results/combinations achieved in Galapagos?

Any help is much appreciated!!!

THANK YOU!

Replies to This Discussion

Permalink Reply by Michal Dengusiak on December 18, 2012 at 5:58am

Hi,

I just run my first simulation and this are my comments and questions and your reply.

(a) you have to allow Evolutionary solver for some time in order to see the first set of results. In your case which seems to be quite complex allow at least one hour

2. I could not find a way of saving results as well, maybe we could try to saving via slider animation no idea will try..

Those are my results.

When I read my Record....I got the values for generation 1 and then for most remaining one I have:

Record: Genome was mutated to avoid collision

Also not sure why Galapagos did not try any combination from a middle left cloud...maybe not enough time to test this option....I will try to leave it running overnight to see how this will evolve...

Permalink Reply by Michal Dengusiak on December 19, 2012 at 2:24am

Hi,

I run it overnight and here is summary:

1. I noticed that is good to make number more round..In my case I did Integer

2. Galapagos has finished after 50 steps and did run simulation for each pool of options

3. However I am not sure that in my example was not better to use Annealing Solver

4. I have in my results 1048567 line most of them are Record: Genome was mutated to avoid collision this mean that it run so many iterations?

Permalink Reply by Michal Dengusiak on December 19, 2012 at 2:26am

Results file

Attachments:

Galapagos_Results.zip, 1 MB

Permalink Reply by Michael Makris on January 22, 2013 at 4:49pm

Hey Michael. I'm new to Galapagos. How does you know how to read your results/get your results from Galapagos? IE how do I learn what all the graphics in that screen mean and how do I save the genome data for analysis?

Permalink Reply by Michal Dengusiak on January 23, 2013 at 12:42am

Hi, In this case above was quite simple. You can see on the left side 4 and right 3 options. Those are the values of sliders - left (percentage of glazing), right (type of glass). I did just simple multiplication of values and looking for minimum. That is what my answer is quite simple and in fact we do not need to run dozens of iteration as enough would be 12 variations and we could get answers. However not sure how to set this up that will just check possible option and then finish?

Normally each slider has different value/importance for us and unit and we shall create a function that would even result and make some nice ie. the logarithmic curve so Galapagos have more space to play. Great description can be found here:

http://ieatbugsforbreakfast.wordpress.com/2011/03/07/define-fitness/

Permalink Reply by Michael Makris on January 28, 2013 at 4:16pm

Hey Michal. Thanks for the reply, but I think you're answering a different question. I'm asking about how to understand the actual Galapagos interface, i.e. the orange graph that Galapagos creates. I'm confused as to what the yellow/orange means as well as the hatched/plain. Essentially, I'm wondering what the y-axis means in the Galapagos graph.

Also, I'm confused as to what the different shades of green mean on the Galapagos genomes in the lower right corner.

Permalink Reply by David Rutten on January 23, 2013 at 1:09am

Record: Genome was mutated to avoid collision

means that a new offspring genome was created which was similar to a genome already once computed (Galapagos never calculates the same genome twice). When this happens a lot it means one of two things:

The entire population is highly clustered in a single spot meaning there's not enough genetic 'room' in between individuals to create unique offspring. When this is the case, it means a (local) solution has been found and there's little point in continuing the process.
The possible gene-pool is very shallow. Say you have 3 sliders, each of which can take on 4 values. That means you have a total of 4³ = 64 possible combinations. The first genome picked is always unique, the second genome (when picked at random) has a ~1.5% chance of being identical. The third genomes (when picked at random) has a ~3.1% chance of colliding with the two existing genomes. And so on. Very soon you'll have exhausted all possible genomes and Grasshopper will frantically try and mutate any additional individual until it becomes unique, but of course never succeeding.

Based on your screenshots the latter is the case. You have two sliders, one with 4 and one with 3 possible states. This gives you a total of 12 possible genomes which corresponds to the 12 lines in the middle graphic and the 12 groups in the left-most graphic. You've computed each genomes more than once and there is no new information to be had.

I suppose Galapagos could be smarter about this and detect when it has exhausted all possibilities, but at the moment it has very little smarts.

David Rutten

david@mcneel.com

Poprad, Slovakia

Permalink Reply by Cheryl Lee on January 23, 2013 at 6:24am

Hi! Sorry, I must have missed all of this.

Regarding the evolutionary solver, leaving it for hours to run did do the trick, especially with the simulations it did take pretty long.

Regarding saving the outputs, I found this which is pretty useful.

http://www.grasshopper3d.com/forum/topics/exporting-galapagos-outputs

Also, you can also try using this (file uploaded) which a friend helped me with. It is very simple to use, the instructions are in the file. He used an additional plug-in called gHowl to export the results into excel. You can download the plug-in here: http://www.grasshopper3d.com/group/ghowl

All credits goes to Thian Siong (not sure what's his screen name).

Attachments:

GalapagosDataRecorderUsingGHowl.gh, 10 KB

Permalink Reply by Rafael Prado Cartana on September 17, 2016 at 11:38am

Hi Cheryl

I'm facing the same problem you describe in this post. I´m doing a UDI (climate based) simulation to a solar shading device (very simple) using Galapagos to optimize it. But as in you case my results doesn't appear on Galapagos screen. I simulated it overnight (12hs) and the problem remains... How did you solve it??? can you help me? Thanks a lot Rafa :-)

Permalink Reply by Rafael Prado Cartana on September 17, 2016 at 11:39am