Reinforcement learning resources for the Performance-oriented Congestion Control project.
This repo contains the gym environment required for training reinforcement learning models used in the PCC project along with the Python module required to run RL models in the PCC UDT codebase found at github.com/PCCProject/PCC-Uspace.
To run training only, go to ./src/gym/, install any missing requirements for stable_solve.py and run that script. By default, this should replicate the model presented in A Reinforcement Learning Perspective on Internet Congestion Control, ICML 2019 [TODO: Hyperlink].
To test models in the real world (i.e., sending real packets into the Linux kernel and out onto a real or emulated network), download and install the PCC UDT code from github.com/PCCProject/PCC-Uspace. Follow the instructions in that repo for using congestion control algorithms with Python modules, and see ./src/gym/online/README.md for detailed instructions on how to load trained models.