"Robert McGregor" <rob### [at] mcgregorfineartcom> wrote:
> Here's a 100x100 unit maze (union of 60,000 boxes), also rendered with an
> orthographic camera.

Pretty cool, Robert. It passes the flood fill test ;) The solution is very
convoluted, though quite linear in the sense that it's easy to determine where
to go. The branches along the way often terminate abruptly or are bounded by the
path already taken. I might have had a problem at the very bottom if I hadn't
known that the exit was on the right wall... but don't feel bad, many of my
previous attempts had obvious solutions too (more so than yours) :)

Care to give an overview of the algorithm?

Last spring I was experimenting with mazes myself, the subject being something I
go back to from time-to-time. I was investigating the use of large point sets
and nearest neighbor searches (using Voronoi diagrams), and using POV as a major
step. The results were satisfactory, but my designs were confined to cases where
the perimeter wall was easy to define, such as circular mazes.

Sam

Post a reply to this message