A1gent-ns3-oran

Agentic ns-3 O-RAN platform: deterministic and auditable intent-driven radio control.

Quick start

1. Create a Python environment and install dependencies:

python3 -m venv .venv
source .venv/bin/activate
pip install -r a1gent/requirements.txt

2. Prepare an ns-3 workspace:

./scripts/bootstrap.sh

Default workspace target: workspace/ns-3.42.

3. Create runtime env files:

cp experiments/configs/ns3-scenario.env.example experiments/configs/ns3-scenario.env
cp experiments/configs/orchestrator.env.example experiments/configs/orchestrator.env

ns3-scenario.env controls the ns-3 scenario. orchestrator.env controls A1gent runtime and policy guardrails.

4. Set OPENROUTER_API_KEY in experiments/configs/orchestrator.env.

5. Run the ns-3 scenario:

./scripts/run_ns3_scenario.sh

6. Run the orchestrator in a second terminal:

./scripts/run_orchestrator.sh

For ns-3-side source changes, see ns3/README.md.

Citation

Please cite:

H. Li, D. Xu, M. Chen, and Y. Liu, "Agentic Open RAN: A Deterministic and Auditable Framework for Intent-Driven Radio Control," IEEE International Conference on Communications (ICC), to appear, 2026.

Acknowledgements

This research was supported by NSF through Award CNS-2440756, CNS-2312138, CNS-2332834, and NVIDIA Academic Grant Program.