OSCAR: a Contention Window Optimization approach using Deep Reinforcement Learning to quickly learn the optimal policies under different network conditions
A working ns3-gym environment is required. Moreover, a wandb account is required to show the results.
Clone this repository so that it lands inside ns3-gym/scratch/linear-mesh.
Edit run = wandb.init(entity="xraulz", project="contention_window", tags = wtags)
in each training/test file and change entity=xraulz
with entity=your_wandb_username
, where your_wand_username
is the wandb account username.
Launch python OSCAR_train.py
to train the OSCAR algorithm.
Launch python CCOD_train.py
to train the CCOD algorithm.
Launch python standard_test_and_ccod_train.py
to test the 802.11 algorithm and train the CCOD algorithm.
@INPROCEEDINGS{10279663,
author={Grasso, Christian and Raftopoulos, Raoul and Schembra, Giovanni},
booktitle={ICC 2023 - IEEE International Conference on Communications},
title={OSCAR: A Contention Window Optimization Approach Using Deep Reinforcement Learning},
year={2023},
volume={},
number={},
pages={459-465},
doi={10.1109/ICC45041.2023.10279663}}