connection through a network of points

Hello to everyone.

I am trying to create a simulation where:

_a number of points moves according to swarm rules. It is restricted by a bounding area, and it is attracted by a number of other points, being perceived as the basic volumes that receive an intensified amount of movements

_then I want to find out where those movements intensify. I have solved it by trying to find the closest points between all the curves that the initial swarm points generate.

_now the issue is that I want to connect those points were the movement is more intense in a way that they form some kind of nest corridors. I have already tried the mesh triangulation component but it does no good, because the number of connections is not stable. Meaning that if a point is close enough with 5 other points I want it to connect with all of them but if the distance threshold includes only one point then it will have to connect with only one point.BRANCHING%20POINTS.PNG

My goal is something like this https://www.youtube.com/watch?v=WOFJ4RAlbPY, but then again any suggestions, concerning workflow or components or optimization, are always more than welcome. I suppose that my result would be much more easy to achieve through scripting, but I would be more than grateful for some further indications.

Attachments:

Replies to This Discussion

Permalink Reply by peter fotiadis on December 30, 2015 at 12:27am

Hmm ... (are you familiar with C#?)

The "by the book" solution is to write down a recursive (fractal) code that does the "tree" that you are looking for. In fact I had exactly that type of stuff but due to the trad summer crash several similar fractal C# scripts (made just for fun) are lost.

A crude solution MAY be by making a graph (= something out of your nodes and connecting edges) and then applying Dijkstra classic shortest path routing. Pics attached demonstrate that (pts are random, connections are random and cost [i.e. values in a classic adjacency Matrix] is the distance from pt to connected pt. BTW:node within the red circle is the target node for the shortest routing from all the rest):

Permalink Reply by Jo Skar on January 1, 2016 at 2:01pm

Thanks for your quick reply, and I wish you a Happy New Year! I am familiar with C# in general but not familiar with the Rhino libraries it includes. The images you have attached are very close to what I am looking for and seem very interesting. Is there any tutorial available in order to start translating the shortest path algorithm in Rhino commands?

Permalink Reply by Jo Skar on January 1, 2016 at 2:22pm

also is it possible that a plugin called SpiderWeb can be of any help?

Permalink Reply by peter fotiadis on January 2, 2016 at 12:13am

Well Jo

1. I have no idea what SpiderWeb does.

2. By "tutorial" for translating some shortest path algorithm (classic Dijkstra or the newer ones that are considerably faster: in case that your graph nodes are quite a few, say some millions) you mean writing the C# code? If so ... I'm not aware of anything similar. But in general you can do it by using the Rhino/GH SDK for accessing mostly the "graphical" part of the equation. Obviously the most sensitive part is the adj Matrix "filling" policy.

Note: if you have plans to use the code outside the GH umbrella I would suggest to make it as GH "neutral" as possible.

Apologies for not publishing the code: it's classified strictly as internal.

best, Peter

Permalink Reply by Jo Skar on January 2, 2016 at 5:28am

Peter,

once again thank you for your quick replies. I understand that you can't publish the code. I want to try and implement it myself in C#,after all doing is the best way to learn. I tried to familiarize myself with the Dijkstra algorithm, and my basic question is how to introduce the points and edges of connection ( geometry) as nodes and graph (data). I hope that I will find my way out of it, unless nodes and graphs are some specific types of parameters that C# uses.

Permalink Reply by peter fotiadis on January 2, 2016 at 6:27am

Well ... nodes are points and connections are lines. The only thing that you must do is fill the adj Matrix based on a connectivity data tree (pt to pt) whilst using some policy for the adjacency matrix (or double[,] or square tree) values ( say: pt to pt distance or some other ??).

Actually an orthodox Dijkstra algo uses double [,] instead of a matrix ... but a matrix is much more suitable for visualization in the GH environment.

BTW: all that stuff IS NOT the best way to start learning C# ... but I have a feeling that you are a brave man > I like that more than anything > therefor I MUST give you a hand (or two, he he).

So finally some good news: ping me a friendship request in order to have my e-mail and then ... we can arrange a live Skype session for all these things (BTW: you can see other ultra freaky things as well ... be prepared, he he).

best

Permalink Reply by Manuel Muehlbauer on November 21, 2016 at 10:04pm