This repository contains research and experiments for network congestion control, centered on the rl-tsc/ pipeline.
The current project direction is:
offline-to-online learning for multi-bottleneck congestion control under topology and traffic shift
The repo includes:
rl-tsc/- main congestion-control pipeline,
- synthetic dataset generation,
- Pantheon trace ingestion,
- behavioral cloning pretraining,
- RL finetuning and evaluation.
images/- project visuals.
read_docx.py- utility for extracting text from the included DOCX file.
Start with: