Open AI-RAN Testbed: One Dashboard for RAN Deployment, RAN Digital Twins and AI-RAN Training
The Open AI-RAN Testbed from TOSSI brings together open-source RAN deployment, radio-environment simulation, machine learning, and AI-driven network optimisation in a unified experimentation platform. From deploying RAN stacks across heterogeneous infrastructure to connecting 3D radio environments with running RAN systems, the platform provides a common workflow for building, evaluating, and operating AI-RAN applications. This release demonstrates how these capabilities work together. What’s covered in this release: Unified RAN Orchestration: Deploy and manage open-source RAN and 5G core configurations through a common automation engine, accessible via the web dashboard and CLI. Open RAN Integrations: Work with configurations involving OCUDU-RAN, OpenAirInterface (OAI), NVIDIA Aerial, xFAPI, and hardware acceleration. Distributed Infrastructure: Manage deployments across servers with different hardware capabilities and match workloads to their infrastructure requirements. Digital Twin with NVIDIA Sionna RT: Build 3D radio environments, model propagation, generate channel data, and integrate simulated channel conditions with OCUDU-RAN. Virginia Tech Campus Scene: Explore the sample Virginia Tech University campus scene included with the platform. This environment will be featured in the Open AI-RAN Testbed demonstration at the OCUDU Developer Summit at Virginia Tech University. AI-RAN Model Lifecycle: Collect data, train and benchmark models, evaluate them in shadow mode, and progress to closed-loop inference. AI-Driven Network Slicing: Explore a spectral-efficiency optimisation use case that dynamically adjusts PRB allocation between eMBB and URLLC slices, aiming to maximise cell spectral efficiency while respecting configured slice SLAs. LLM-Based Operator Agents: Interact with platform operations through natural language, with approval controls for agent-driven actions. The central idea is to connect these capabilities within one engineering environment. RAN deployment, digital-twin simulation, data collection, model training, and runtime evaluation become parts of a common workflow rather than isolated experiments. The Open AI-RAN Testbed provides a foundation for reproducible experimentation and continued development of open-source, AI-driven radio access networks. Documentation: Explore more from TOSSI: