Code containing RRM simulation using RL in a scenario with RAN slicing.
- Install pipenv
- Install dependencies using pipenv:
pipenv install - To access the virtual environment created, run
pipenv shell, now all commands which you run will be performed into virtual enviroment created - (In case you want to contribute with this repo, if not you can skip this step) Activate pre-commit hooks to use black formatter, flake8 lint and Isort references. Run
pre-commit install. Now every time you make a commit, black formatter, flake8 and isort will make tests to verify if your code is following the patterns (you can adapt your IDE or text editor to follow this patterns, e.g. vs code)
Run the script optimize.py using pipenv running pipenv run python optimize.py. It would take a long time to Optuna generate the optimized hyperparameters for each scenario, so you can use the hyperparameters that it were already generated into hyperparameter_opt folder and skip this step.
Run the command pipenv run python run.py to start the simulation for all scenarios. You can watch the training performance using tensorboard. After the training finish, the RL models will be saved into agents folder, the VecNormalize parameters into vecnormalize_models folder, the evaluations made along with the trainign into evaluations folder, and the test results into hist folder.
To generate the figures with results obtained in the paper, you can run pipenv run python plot_results.py and the figures should be available into the folder `results" as soon as the script finish.
To cite this repository in publications:
@ARTICLE{nahum2023rrs,
author={Nahum, Cleverson V. and Lopes, Victor Hugo and Dreifuerst, Ryan M. and Batista, Pedro and Correa, Ilan and Cardoso, Kleber V. and Klautau, Aldebaro and Heath, Robert W.},
journal={IEEE Transactions on Wireless Communications},
title={Intent-aware Radio Resource Scheduling in a RAN Slicing Scenario using Reinforcement Learning},
year={2023},
volume={},
number={},
pages={1-1},
doi={10.1109/TWC.2023.3297014}}