python3 -m venv pyRLLander
source pyRLLander/bin/activate

wheel
Box2D (box2d-py or box2d-kengz on Mac)
pyglet
wandb
scipy

First, clone the OpenAI Gym repository: https://github.com/openai/gym

mv lunar_lander.py ../gym/gym/envs/box2d/
cd ../gym
pip install -e .

In the multi_agents.py file, choose your starting configuration. Then, run the following command:

python multi_agents.py

In test.py, update the location of the agent's pytorch file and the envionment paramters file.

python test.py

I trained the model on 5 independent environments with different parameters.

The following plots show how the elite agent performed as it worked towards solving the environment.

• POET (https://arxiv.org/abs/1901.01753)
• Self-Constructing Neural Networks Through Random Mutation (https://arxiv.org/abs/2103.15692)
• OpenAI Gym (https://github.com/openai/gym)
• Weights & Biases (https://wandb.ai)