Snake AI

An AI based on a genetic evolutionary algorithm, that learns how to play snake on a 10x10 matrix. For visualizing the AI, the best snake overall (number of eaten snaks) is captured in an mp4 file called "Fig.mp4"

Arhitecture:

Input:

1. Xfood - Xhead
2. Xhead - Xfood
3. Yfood - Yhead
4. Yhead - Yfood
5. distance to wall/tail down
6. distance to wall/tail right
7. distance to wall/tail up
8. distance to wall/tail left

Output:

1. probability of going down
2. probability of going right
3. probability of going up
4. probability of going left

Decision:

Making the snake making one step towards the dominant decision from Output

Processing:

The exact value for the constants is given in the file 'GlobalConstants.py'.

Each generation has N individuals, playing on a DIM x DIM matrix, and having at most T time stamps without eating to make the biggest fitness.

After each generation, the best SURVIVAL snakes evolve in N / SURVIVAL new snakes.

Fitness:

The fitness of a snake is the number of times it ate.

Evolution:

Each individual evolves into N / SURVIVAL new individuals.
The evolution process is the following:
1. Each snake breakes down into 2 copies: one identical and one with some minor mutations.
2. Pairs of snakes cross-bread: one new snake is made with a part of the ADN of the first snake and a part of the second one, WITHOUT mutations.