[leni536]

I assume Neumann 0 boundary condition at the walls. Would it always solve the maze though? It seems intuitively true but hard to prove.

[s-macke]

Hacked something together. Seems to work.

http://simulationcorner.net/maze/

[s-macke]

Yes, open boundary conditions at the walls.

I am not sure what happens if you have two possible paths to the exit but one of them is wider, but longer. It might not always find the shortest path.

I like this main idea, because it is independent on the dimension and structure of the maze.