# OpenRAN Gym: An Open Toolbox for Data Collection and Experimentation with AI in O-RAN

arXiv: https://arxiv.org/abs/2202.10318 · HTML: https://mlsanity.com/paper/2202.10318/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 Feb 2022.

## Built on by (most cited first, 7 of 7)

- [Programmable and Customized Intelligence for Traffic Steering in 5G Networks Using Open RAN Architectures](https://mlsanity.com/paper/2209.14171/) (2022)
- [OpenRAN Gym: AI/ML Development, Data Collection, and Testing for O-RAN on PAWR Platforms](https://mlsanity.com/paper/2207.12362/) (2022)
- [Actor-Critic Network for O-RAN Resource Allocation: xApp Design, Deployment, and Analysis](https://mlsanity.com/paper/2210.04604/) (2022)
- [Adaptive Resource Allocation for Virtualized Base Stations in O-RAN with Online Learning](https://mlsanity.com/paper/2309.01730/) (2023)
- [Intelligent Closed-loop RAN Control with xApps in OpenRAN Gym](https://mlsanity.com/paper/2208.14877/) (2022)
- [FORLORN: A Framework for Comparing Offline Methods and Reinforcement Learning for Optimization of RAN Parameters](https://mlsanity.com/paper/2209.13540/) (2022)
- [Z2-ACT: End-to-End Verifiable Agentic Intent Control for Open 6G RAN](https://mlsanity.com/paper/2608.21049/) (2026)

## Abstract (arXiv)

Open Radio Access Network (RAN) architectures will enable interoperability, openness, and programmatic data-driven control in next generation cellular networks. However, developing scalable and efficient data-driven algorithms that can generalize across diverse deployments and optimize RAN performance is a complex feat, largely unaddressed as of today. Specifically, the ability to design efficient data-driven algorithms for network control and inference requires at a minimum (i) access to large, rich, and heterogeneous datasets; (ii) testing at scale in controlled but realistic environments, and (iii) software pipelines to automate data collection and experimentation. To facilitate these tasks, in this paper we propose OpenRAN Gym, a practical, open, experimental toolbox that provides end-to-end design, data collection, and testing workflows for intelligent control in next generation Open RAN systems. OpenRAN Gym builds on software frameworks for the collection of large datasets and RAN control, and on a lightweight O-RAN environment for experimental wireless platforms. We first provide an overview of OpenRAN Gym and then describe how it can be used to collect data, to design and train artificial intelligence and machine learning-based O-RAN applications (xApps), and to test xApps on a softwarized RAN. Then, we provide an example of two xApps designed with OpenRAN Gym and used to control a large-scale network with 7 base stations and 42 users deployed on the Colosseum testbed. OpenRAN Gym and its software components are open source and publicly-available to the research community.