Use deep Q-learning to train a model to play the game snake. Projects of this scale quickly remind me of why I hate Python for substantial projects.
The game is given as a grid where the upper left corner is (0, 0). Elements are access by (x, y) positions,
corresponding to column, row positions.
- The snake may not move into a self-intersecting position.
- The snake does not grow immediately upon eating. Rather, the tail must reach the position where the food was eaten before it actually grows.
- By default, the snake dies when it hits a wall.
- Food is removed instantly once the snake makes contact with it.
The main method for the text based game is, fittingly enough, located in snake_ai/text_game.py. A version with real
graphical rendering may be done with pygame at a later time.
All tests are found in the test directory. Nose is probably the best way to
go about running them, but any testing framework capable of running Python's standard unit tests should work.