Skip to main content

Hello,

I was asked something today, where I'm not sure. Maybe?

 

Someone wondered if I can make a FME workflow to evaluate possible design options for some room layouts. The keyword they used was "generative design". It's a stupid marketing term, pushed by Autodesk. So let's call it "what if analysis".

 

Main focus would be placing tables to optimise views, distance to elevator cores, etc. All very doable in 2D...

 

6a00d83452464869e20240a4e802ce200bSo, I wonder... can I? Did someone attempted something like that?

I'm thinking Monte-Carlo, or a simple neural network in R, or Python?

 

I'm aware, there are quite a few tools that do this out of the box. The advantage I see with FME is that this would save me a lot of trouble parsing geometry, much simpler as a workflow.

Interesting problem! I'm quite sure you could do it in FME, but I'm not sure it's the most intuitive tool for something like this. But I would definitely consider FME as part of a solution, perhaps to "glue" different components together.

Personally I'd start by looking into Python and algorihtms related to bin-packing, although you'll probably need to add parameters to prioritize the different ways to optimize the spaces (views, walking distance, min/max group sizes, etc.).

However, I suspect that the problem is actually fairly hard (if not NP-hard)...


I seem to recall a presentation from GIM a few years ago, I think it was @bruno_de_lat​, who had used the MapTextLabeller to determine the best location of a tower crane on a building site. It'd be a start I suppose.


Reply