Using the files from Kirkados' JSR2020_D4PG as a base for DRL-based guidance of spacecraft proximity operations trained via D4PG.
Set-up coding environment as described below. This has been tested using Python 3.6.8 on Windows10 and Python 3.6.13 in a Conda environment on Ubuntu 20.04.6 LTS.
pip install tensorflow-gpu==1.12.0
pip install tensorflow==1.12.0
pip install protobuf==3.19.6
pip install psutil
pip install pyvirtualdisplay
pip install scipy==1.5.2
pip install matplotlib
pip uninstall numpy
pip install numpy==1.16.4
pip install ffmpeg-python
pip install shapely==1.8.5.post1
In Windows: extract, add bin folder to PATH: https://github.com/GyanD/codexffmpeg/releases/tag/2022-03-28-git-5ee198f9aa
To train a model, modify settings in environment_envs123456.py and settings.py. Then run
Windows:
py main.py
Ubuntu:
python main.py