Published April 20, 2026 | Version v1
Software Open
Cooperative Deep Reinforcement Learning for Fair RIS Allocation
- 1.
TU Wien
- 1. TU Wien
Description
# RIS Gambling
This project implements and evaluates reinforcement learning (RL) and heuristic methods for Reconfigurable Intelligent Surface (RIS) allocation in wireless networks.
The paper describes the purpose of the research in greater detail, as well as the implementation, especially in Subsections IV/D and at the beginning of Section V. The included codebase and trained RL models make the results presented in the paper reproducible. The codebase can be used by running the core modules from the CLI folder (described later), and the contents of the src folder can also be reused in other projects, as it contains the environment and several wireless network–related components (such as beamforming and geometry generation).
Core modules:
- `cli/train.py` – train and save RL models.
- `cli/evaluate.py` – run experiments and compare RL vs. heuristic baselines.
- `cli/plots.ipynb` – visualize performance metrics.
- `src/config.py` – configurate the environment, training and testing.
- `src/` – configuration, environment, metrics, and utility functions (e.g., plotting).
Python version: 3.12.7
Licenses:
Data: CC BY license
Software: MIT license
Files
Code and results.zip
Additional details
Related works
- Is supplement to
- Publication: 10.48550/arXiv.2603.25572 (DOI)
Funding
- FWF Austrian Science Fund