-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Welcome to the ai-snake-simulator wiki!
The goal of this project was to implement the Q-Learning algorithm for teaching an AI to play the snake game! This game provides a different kind of challenge as opposed to the QL racecar-simulator since the AI will be taking in the entire board as a sequence of inputs, meaning that the AI will have to discern which direction the snake is moving.
The simulator itself was meant to be simplistic but allow for expansion if the desire arises to make things more complicated. The snake will take in two frames of the game-board at a time, where each pixel will have differing values depending on what is in the cell:
FOOD_CELL = .75
SNAKE_CELL = .25
NOTHING = 0
Learning from my past mistakes within the ai-racecar-simulator project, I wanted to separate the simulator from the model of the simulator, so that the UI is completely unnecessary. This will allow me to better make use of higher processing power (using a super computer) or use a grid search for best parameters.