Congestion control using deep reinforcement learning with ns3
This was a part of my QSIURP research project in summer 2024. We investigated the landscape of machine learning application to improve system performance as a whole and delved deeper into using deep reinforcement learning for congestion control in particular. Using ns3 for data collection was a direction that few prior work had picked up, despite its potential.
- Anaconda
- ns3
Create and activate the environment from the saved specs:
conda env create -f env.yml
conda activate qsiurp_env
Install ns3-gym as instructed. Then clone this repo in the examples folder of opengym.
Start training by running the following command:
python train.py
Evaluate and plot the graphs by changing LOG_DIR and MODEL_FILE and run eval.py:
python eval.py
This project is licensed under the MIT License - see LICENSE for details