Skip to content

This is python project aimed at evolving creatures that are simualted in 2D world where, the goal is to identify such Neural Nets that perform best for the given fitness function using genetic algorithm.

Notifications You must be signed in to change notification settings

AyushBobale/NeuroEvolutionUsingNEAT

Repository files navigation

NeuroEvolutionUsingNEAT

A python project created to train creatures how to walk as far as possible in a single direction.

The 2D world is simulated using the pymunk library of python.

The genetic algorithm is implemented using the python's NEAT library.

New Scrollable Environment

ScrollabelEnvironment

To Move camera

  • Left Arrow key
  • Right Arrow Key

New Organism Bipedal

For initial generations

Starting Generation

For later generations [Not what I expected to be the behaviour even with a custom modified fitness function]

Ending Generation

Best genome pickling and loading for test replays added

How to run

git clone https://github.com/AyushBobale/NeuroEvolutionUsingNEAT
cd NeuroEvolutionUsingNEAT
pip install pymunk
pip install pygame
pip install neat-python
python main.py 

Previous Work

Sample Organism

Simulation Screen Shot

For initial generations [Generation 1]

Starting Generation

For later generations [Generation 81]

Ending Generation

About

This is python project aimed at evolving creatures that are simualted in 2D world where, the goal is to identify such Neural Nets that perform best for the given fitness function using genetic algorithm.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages