Before I talk fluids, I first want to catch up on my Pac-Man 256 clone progress. At first I tried to look into ways to procedurally generate the maze the player navigates using traditional maze generating algorithms. After playing around with a modified breadth-first search approach (the results of which can be seen below), I realized the results looked a little too random and were a far cry from the levels found in the game.
Upon playing the original game a few more times I started to notice a few consistent things about the board. This included things like the starting layout always being the same:
and a seam appearing every so often:
This suggested that the endless levels were more likely comprised of premade sections that are stitched together at predictable seams. Because of this, I switched my approach from purely random generation to stitching together of premade sections. So far, I’ve managed to put together a simple editor for designing the chunks to be stitched together. For example, a Mario of sorts:
Next up is exporting the designed sections and then developing a system for continuously loading them into the game.