A script that lets you play Lunar Lander with one of our agents assisting you.
Press "W" to fire the main engine, "A" to fire the left engine, and "D" to fire the right engine.
A red box in the top right corner indicates that the agent is overriding your action.
In general, to run keyboard_agent.py requires:
- A Python 3.8 environment
python -m venv ".\env"
.\env\Scripts\activate
python -m pip install --upgrade pip setuptools wheel
- PyTorch 1.8+, such as with this command
pip3 install torch==1.8.1+cpu torchvision==0.9.1+cpu torchaudio===0.8.1 -f https://download.pytorch.org/whl/torch_stable.html
- Install all the packages in
requirements.txt
pip install -r requirements.txt
For Windows, we include a batch script to install the environment for you, assuming you have Python 3 on your path available as python.
To install the Python environment in the current directory run
create_env.bat
From there as long as the environment is activated (with .\env\Scripts\activate) you can run the command in the below section to play the game)
To play Lunar Lander with our agents, you can run the following command
python keyboard_agent.py --trained_agent_type 1 --zoom_level 2
trained_agent_type determines which autonomous agent will be assisting you.
- 0: no assistance (vanilla lunar lander)
- 1: the "best" agent, trained with a "sensor" human and intervention penalty of 1
- 2: trained with a "noisy" human and intervention penalty of 0.15
- 3: trained with "noisy" human and intervention penalty of 0.75
- 4: Ensemble of 1, 2, and 3. i.e. an action is sampled uniformly randomly from one of those agents at each timestep
Adjust zoom_level if the image is too small or too big (smaller number means smaller image)
NOTE: If you reach the goal, the game might not reset for a while, so just re-run the script in this case.