Connectivity Optimization

Good morning,

While this is certainly not a Grasshopper question I reckon that people into parametric design also have a knack for picking up funky algorithms.

Do anyone know if there is a know algorithm for sorting connectivity between nodes. More specific to find the most efficient node layout without connection overlaps (or the fewest). For an example:

I'm thinking of using it to sort/evaluate architectural programs, but don't even know what to look for.

/Cheers

Replies to This Discussion

Permalink Reply by robert cervellione on March 30, 2010 at 10:12pm

there is a plugin for rhino called point set reconstruction. in it is a tool called "network". try that out

Permalink Reply by Mattias Eriksson on March 31, 2010 at 3:29am

yeah, unfortunately it deals with finding the closest neighbours, where I want to relocate the point to eliminate intersections. After googling around for a couple of hours I can't even find anyone that tried.. maybe it's just more or less impossible

Permalink Reply by Luis Fraguada on March 31, 2010 at 4:14am

delaunay triangulation? that will get you part of the way there perhaps...

Permalink Reply by robert cervellione on March 31, 2010 at 8:31am

i now understand what you are trying to do. i did a project where i scripted springs to distribute program in 3d based on programmatic relationships/factors. see the little test video below. it was done in maya but i assume if kangaroo lets you create springs between elements then you could get the same effect in grasshopper. in my case intersecting lines was not a issue because the lines(springs) represented a the relationship from one element to another (like adjacency factor to each other as well as site factors).

-robert

Attachments:

progTest.f4v, 3.2 MB

Permalink Reply by enrique soriano on March 31, 2010 at 4:15am

Hi Mattias

According to your drawing I can imagine a very complex problem, and I think a Front advancing Technique, will probably not solve it.

I see 2 options:
-relaxing the whole network (but u'll need to take care of the boundary) with an repulsion-atraction iteration. (force-based diagrams are very used to visualize graphs)
-if the network is really messy,( and the first option gets quickly stucked, or bugs) I guess you could evaluate better results (probably not the best) throught GA .

but probably I would consider to especify more clearly what kind of network topology you have,
I mean, maybe if you have a large drwaing.. you can describe better the problem to solve.

Permalink Reply by David Rutten on March 31, 2010 at 4:55am

I'd definitely start with option 1. There's no guarantee that it will solve link crossings that could have been solved, but it tends to give good results.

I attached three videos I made years ago (probably in 2005) that show a very naive force-density network relaxation.

All nodes repel each other, connected nodes also attract each other.

--
David Rutten
david@mcneel.com
Poprad, Slovakia

Attachments:

Permalink Reply by Daniel Piker on March 31, 2010 at 5:14am

Kangaroo could be used for this sort of approach.
Beta launch later today.
kangaroophysics.com

Permalink Reply by Pep Tornabell on March 31, 2010 at 6:01am

Amazing news Daniel

Cool stuff David

Interesting theme all!

Permalink Reply by enrique soriano on March 31, 2010 at 6:04am

"Beta launch later today."
yeah.!
Great expectations
Yummy holydays!

Permalink Reply by Mattias Eriksson on April 5, 2010 at 9:25am

Yay, exicited as I am about Kangaroo (which I am) I'm starting to realize just how involved this problem is. I'll look into all your suggestions over easter.

Happy holidays!

Permalink Reply by paul hackett on April 13, 2010 at 11:09am

Any ideas of how can we get this dynamic network with kangaroo? It seems quite complicated to create dynamic distance constraints . David, can you please share your network relaxation definitions?

RSS