Adaptive Contention Window Design using Deep Q-learning

This repository is the official implementation of "Adaptive Contention Window Design using Deep Q-learning".

Requirements

To install requirements:

docker pull pytorch/pytorch

📋 (Optional) Install NS3 to generate new simulation dataset. Dataset used in the paper for training is provided in the Dataset directory.

Training

To train the model(s) in the paper, run this command:

python trainRL.py  --n 10 --ps 1 --transitionModel Markovian --history 0

n: wireless network size; {5,10,20}

ps: transition probability of stochastic process; {0.75,0.9,1}

transitionModel: stochastic model followed by other nodes; {'Markovian', 'NonMarkovian'} (Here 'NonMarkovian' corresponds to the 'complex' process in the paper.)

history: number of previous time steps from which observations are used as input; {0,1,2,3}

Evaluation

To evaluate the model, run:

python eval.py --n 10 --ps 1 --transitionModel Markovian --history 0

Pre-trained Models

📋 Pretrained models are present in modelRL directory.