Version: Python 3.7
AI to simlulate and play the game snake in attempt achieve maximum score.
The Snake uses the "Q-learning" Reinforcement learning technique to best pick it's next move.
The Snake does not play perfectly, this is due to the limits of space storage in basic Q-learning.
The ammount of storage required is given by States x Actions.
The snake has a limited number of sensors:
Food:
Above, Below, or Inline
Left, Right, or Inline
Danger:
Ahead, Left, or Right
Snake:
Facing Direction
A limit number of moves:
Turn Left, Turn Right, Go Forward
This provides a total of 288 possible states. The addition of extra states causes exponential growth.
Running the existing provided Q-Table.
$ python snakeAI.pyOverriding the existing Q-Table with new values and no gameplay display.
$ python snakeAI.py --learn 100000 --no-display