[reneherse]

I too am interested to know if anyone has experimented with these ideas in a virtual environment.

It seems there would also be an opportunity for enriching procedurally generated worlds with these concepts of landscape (and shelter) patterning.

And not only Olmsted's ideas. Though I'm not up on the state of the art, a couple other useful landscape and architectural theories come to mind:

-Jay Appleton's "Prospect and Refuge" theory of landscape painting analysis (and its extension into architecture by Grant Hildebrand) [1]

-Chris Alexander's "A Pattern Language" and other books.

I wonder how one would go about training a model using data from existing environments, paintings and photography that exhibit these patterns?

[1] https://www.researchgate.net/publication/286123657_Prospect_...