This gym leverages NS3 and WebRTC, which can be used by reinforcement learning or other methods to build a Bandwidth Controller for WebRTC.

You can use this Gym by a Python interface that was defined in gym.py. Here is an example gym-example to use this Gym training a bandwidth estimator.

git clone https://github.com/OpenNetLab/gym gym
cd gym

sudo apt install libzmq5 python3 python3-pip
python3 -m pip install -r requirements.txt
# Install Docker
curl -fsSL get.docker.com -o get-docker.sh
sudo sh get-docker.sh
sudo usermod -aG docker ${USER}

If your OS is ubuntu18.04 or ubuntu20.04, we recommend you directly downloading pre-compiled binary, and please skip the step Build Gym binary

The pre-compiled binary can be found from the latest GithubRelease. Please download and uncompress it in the current folder.

wget https://github.com/OpenNetLab/gym/releases/latest/download/target.tar.gz
tar -xvzf target.tar.gz

make init
make sync
make gym # build_profile=debug

If you want to build the debug version, try make gym build_profile=debug

python3 -m pytest alphartc_gym

Thanks SoonyangZhang provides the inspiration for the gym