generative housing model

Comment

Comment by Marc on January 11, 2014 at 5:24am

Wow, thanks so much. I'm really looking forward to your comment about the rules!
Great work so far!

Comment by Renee Puusepp on January 11, 2014 at 4:31am

Hi Marc. Thanks!

You are absolutely right - I am using cell centres for the particle movement, as you put it. It is a bit like Reynold's flocking model i.e. all particles or agents consider other agents within a certain radius only. Closest neighbour's are simply defined by euclidean distance. 

The tricky bit was to invent the movement rules that cause agents to line up or produce T, H and L shapes. I will write about these rules in more detail soon - hopefully in the next publication. Will post an update here when published.

Comment by Marc on January 9, 2014 at 3:48am

Hey, Renée. Amazing generative work!

Could you please explain the basics of how you manipulate the geometry to find neighbours.
Since you wrote that you have an underlying particle system, I guess that each cell's center is the current particle location, but how you evaluate the relations to the neighbours that produce combinations of geometrical units?
Do you depart from a basic unit or how do you achieve this?

Your work is astonishing and I hope that you are willing to share some of your techniques? :)

Comment by Artyom Maxim on January 6, 2014 at 7:20am

Thanks for the reply,

I've tried using Hoopsnake, it's really very usefull sometimes. I do have to dig into the Python though, especially with their last release(ability to multithread any GH component).

Anyways, keep up the good work!

Comment by Renee on January 6, 2014 at 5:14am

Thanks, Artyom! Your study looks very cool.

I have used an agent based approach where each unit (a cell or a dwelling, if you wish) is responsible for finding its own location. The agent movement is coded in a custom Python component, the geometry is created with standard GH components, and finally the solution is analysed with DIVA. This all is closed in a single loop with the Hoopsnake component (have a look, if you haven't yet - it may help you too). 

So basically, each agent/unit keeps moving around until it finds a place where it gets enough direct sunlight. The resultant layout is thus completely emergent and no human intervention or input besides the number of units is needed. 

Comment by Artyom Maxim on January 5, 2014 at 1:40pm

Very interesting!

I've been experimenting with something relatively similar a while ago, I used Kangaroo in a feedback loop setup with Ghowl to give responsiveness to the predefined areas.

http://www.grasshopper3d.com/photo/urban-zoning-1?xg_source=activity

What workflow did you use?

Comment by Renee Puusepp on January 5, 2014 at 4:21am

The story is published on http://www.reneepuusepp.com/generative-housing-model/

Comment by LLXXZZ on January 3, 2014 at 5:03pm

i cannot understand.