ib-ORAN: Intent-Based Open RAN

August 1, 2026 · View on GitHub

An ns-3-based Open RAN simulation environment with native support for intent-based orchestration.

License

The original contributions in this repository are licensed under the MIT License.

This means you are free to use, modify, and distribute this software, with or without modification, provided you retain the original copyright notice. See the LICENSE file for the full terms.

This project makes use of third-party open-source components. Their copyright notices and license terms are preserved in their respective directories and summarized in THIRD-PARTY.txt. Do not remove or modify the license and copyright files present in any third-party component directory.

Overview
Architecture
Configuration
Getting Started
Troubleshooting
Contributing
Acknowledgements
Further Reading

Overview

ib-ORAN (module name: ns3-iboran) is a comprehensive simulation environment and design framework built on the proven ns-3 network simulator. It is primarily developed to implement and verify intent-based operations within Open RAN architectures.

This framework enables direct control of Physical Layer (PHY) and Medium Access Control (MAC) functions through RAN applications — rApps and xApps hosted on RAN Intelligent Controllers (RICs), as well as distributed dApps — spanning the full spectrum of control timescales from non-real-time (non-RT) to near-real-time (near-RT) to real-time.

ib-ORAN bridges high-performance ns-3 C++ simulations and Python-based policy logic via a shared-memory interface, decoupling the simulation engine from the control plane without sacrificing performance. This architecture natively supports non-RT and near-RT RIC implementations alongside real-time dApp design, enabling intent-based, semantics-aware exchange of payload data across diverse use cases.

Building on the foundations of ns3-gym and ns3-ai, ib-ORAN extends both frameworks with intent-based and semantics-aware capabilities purpose-built for Open RAN network control — going beyond reinforcement learning hooks to provide a full orchestration layer for intelligent, policy-driven RAN behavior.

ib-ORAN supports both 3GPP LTE and New Radio (NR) standards:

LTE: implemented through an adapted version of the built-in lte module in ns-3.
NR: built on top of the ns-3 nr module.

Architecture

The architecture spans three control planes — non-RT RIC, near-RT RIC, and real-time dApp — interconnected via the standardized E2 and A1 interfaces, plus the newly introduced E3 interface, which is not yet standardized. The shared-memory bridge between ns-3 and the Python orchestration layer is a core design principle, enabling clean separation of simulation logic from intelligent control.

Concretely, A1 carries policy guidance from the non-RT RIC's rApps to the near-RT RIC's xApps; E2 connects the near-RT RIC to the O-DU (ns-3 RAN), carrying KPM reports, RIC control, and dApp-related messages; and E3 connects the dApp to the MAC/RLC layers, enabling sub-10ms local control at the O-DU.

Message Structures

See the message-structure.md for the complete lists of control messages by service model exchanged over the implemented interfaces.

App Inventory (dApp / xApp / rApp)

Browse the app-catalog.md for the full list of developed dApps, xApps, and rApps. To contribute a new app or algorithm, refer to the CONTRIBUTING.md.

Configuration

Simulation Data

Browse the simulation-data.md for the full details of simulator input and output data, plus the supporting datasets and embedding extraction methods.
Default simulation configuration parameters are defined in iboranConfig.yaml.

Intent Production

ib-ORAN produces intents as sets of object IDs — one set per link direction — requested by, or of interest to, a network operator. The default intent production method is:

Pseudo-Random — samples a fixed number of object IDs per link direction uniformly at random from --oid_range, seeded by --k for reproducibility.

Note: More advanced intent production methods (specifying object IDs directly instead of sampling them) are currently under development.

Getting Started

Installation and Testing

A structured installation and testing guide is available in the install-and-test.md.

Performance Evaluation

Beyond single runs, the iboran_campaign.py script facilitates large-scale evaluation by sweeping over different parameters and algorithms. This makes it easy to compare RAN operation strategies from multiple perspectives using interactive computing platforms (e.g., Jupyter Notebook).

See iboran_evaluation.ipynb for an illustrative example showing how to set up comparisons across key parameters and scheduling algorithms.

Visualization via NetAnim

ib-ORAN automatically generates an XML animation file compatible with NetAnim, enabling graphical visualization of nodes, packet transmissions, and mobility patterns.

Step 1: Clone and Build NetAnim

cd /path/to/ns-3
git clone https://gitlab.com/nsnam/netanim.git
cd netanim
mkdir build && cd build
cmake .. && cmake --build .

Step 2: Launch NetAnim

cd build
./netanim

To display the animation, launch the generated XML file (iboran-lte-anim.xml or iboran-nr-anim.xml) from iboran/notebook/netanim/.

Troubleshooting

Python Module Import Errors

Verify that Python dependencies are installed in the correct environment:

pip install -e contrib/ai/python_utils
pip install -e contrib/ai/model/gym-interface/py

@software{iboran,
  author    = {Agheli, Pouya and Lefebvre, Grégoire},
  title     = {ib-ORAN: Intent-Based Open RAN},
  year      = {2026},
  url       = {https://github.com/Orange-OpenSource/ns3-iboran},
  license   = {MIT}
}

Or the corresponding publication available at IEEE Xplore and arXiv.

ib-ORAN: Intent-Based Open RAN

License

Third-Party Components

Table of Contents