Here is a little neural network simulation meant to work within Pico-8s memory. The network uses integers between 0 and 255 instead of decimals so each network can be easily saved and loaded from cartridge ROM.
You can make a "tank", give it inputs that act as sight, and decide how big the neural network is. Then you can put these new tanks into a map and have them "learn" (not really).
UI and UX are absolute garbage so I apologize in advance.
The controls for the sight editor are: Right click and hold to move a cone. Scroll wheel to change the size of a cone.
During the simulation, you can use the pause menu to fast-forward or spawn a player-controlled tank. The Z/O button changes the camera's current target. Use the pause menu>edit tank to return to the main menu and edit or save.
[Please log in to post a comment]
