Snake-NEAT

A self-learning snake game implemented using NEAT algorithm

About

Project developed at the college as a final work for the artificial intelligence subject

Original code developed between May 15, 2018 - Jun 29, 2018

How to run

Install dependencies

pip install -r requirements.txt

Run

python game.py

Dependencies

• Pygame: Used in game development and user interaction
• neat-python: Python implementation of the NEAT neuroevolution algorithm

Controls

• F1: Takes control of the game, making it possible to control the snake's movements with the keyboard arrows
• F2: Sets a high frame rate (fast) until the snake dies
• F3: Sets a low frame rate (slow) until the snake dies
• F4: Enable the debug draw, drawing input indicators

Fitness function

SnakeLength - TurnedQuantity * 0.01

• SnakeLength: Snake length defined by how many apples it ate

• TurnedQuantity: How many times did the snake change direction, either left or right

Inputs

1. cd_left: Collision-free units at left
2. cd_top: Collision-free units at top
3. cd_right: Collision-free units at right
4. cd_bottom: Collision-free units at bottom
5. cd_top_left: Collision-free units at top left
6. cd_top_right: Collision-free units at top right
7. cd_bottom_left: Collision-free units at bottom left
8. cd_bottom_right: Collision-free units at bottom right
9. apple_x_distance: If there is an apple on the right or left, this variable contains the distance in units to the apple
10. apple_y_distance: If there is an apple on the top or bottom, this variable contains the distance in units to the apple

You can "see" these inputs in debug mode

Outputs

• 0: Don't change direction
• 1: Turn right
• 2: Turn left