Game of Life

Simulates a small world with organisms that can eat food, eat eachother, and make offspring. Decisions of the organisms are based off a neural network based on the genome of the organism. The genome corresponds to the hyperparameters of the network.

How to Run

python game.py

Get argument options

python game.py -h

Generate training data

python game.py --random --store-data --no-gui --games=15 --no-memory-read

Play using genomes + neural networks and then replay it after

python game.py --max-ids-to-read=50 --no-gui --store-history
python play_game_log.py -i data/history_log.csv -a data/history_arguments.csv

Basic game info

Organisms

• Have genome that dictates behavior and traits
• Can either eat food particles or other creatures
• Can mate with other organisms and produce offspring if in adjacent tile
• Starts off with X amount of energy
• Has X amount of maximum amount of energy
• Gain energy by eating food particles or organism
• Each turn expends 1 energy
• Actions:
  • Move
    • Can move north, east, south, or west
  • Eat
    • Can eat a food particle if adjacent tile (X amount of energy)
    • Can eat another organism if adjacent tile (X amount of energy)
  • Mate
    • Can mate with another organism if adjacent tile
    • Spawns a baby organism with the combined genome of the parents
    • Baby placed randomly on map

Food particles

• Static
• Randomly placed throughout the world and slowly regenerated
• How does differing the rate of regeneration change behavior?

TODOs

• Better storage of history log + arguments and allow for different naming schema