RLinWiFi-avg-queue-level

This repository presents a contention window optimization solution for Wi-Fi where the network information is based on the averaged normalized transmission queues’ level.

Prerequisites

1. Installation of python 3.6 and tensorflow dependencies are needed to run the simulation.

   Installation of the following libraries are also needed: tqdm==4.35.0 torch==0.4.1 torchvision==0.2.1 tensorflow==1.14.0 numpy==1.16.3 comet-ml==2.0.12

2. Installation of the ns3-gym environment (https://github.com/tkn-tub/ns3-gym).

3. CometML account is also needed (https://www.comet.ml/signup) to view the graphical results. After creating it update the comet_token.json file with your credentials.

Run BEB tests

positional arguments: N number of stations for the scenario (min: 5)

optional arguments: -h, --help show this help message and exit --scenario SCENARIOS [SCENARIOS ...] scenarios to run (available: [basic, convergence]) --beb run 802.11 default instead of look-up table

Example:
```bash
python agent_training.py                                          # DDPG agent
python tf_agent_training.py                                       # DQN agent
python beb_tests.py --beb 5 10 15 --scenario basic convergence    # Original 802.11 backoff