Researchers at Northeastern University have demonstrated a new open-source AI-RAN prototype that integrates hardware and software from multiple technology providers.
This development brings together components from AmpliTech Group, OpenAirInterface, and Nvidia into a single standards-compliant platform. Consequently, the system shows that open software can manage complex wireless operations.
Technical Architecture of the System
The setup combines AmpliTech’s 64-antenna O-RAN Category B radio unit with Nvidia’s AI Aerial software. Specifically, the Nvidia software handles Layer 1 and Layer 2 processing for the radio access network. Meanwhile, the system uses the OpenAirInterface L2+ stack to complete the software suite.
At the core of this setup is a two-stage precoding architecture. This architecture pairs a four-layer MIMO precoder with a 64-antenna element codebook-based beamformer. Furthermore, the MIMO precoder runs on Nvidia’s GPU-accelerated platform to process signals efficiently.
The Role of open-source AI-RAN
The demonstration shows that an open-source AI-RAN architecture can achieve high performance without relying on proprietary, closed-stack solutions. Historically, massive MIMO systems required tightly integrated vendor implementations. However, this new platform proves that interoperable components can work together.
Testing and Validation Results
The testing took place at Northeastern’s Institute for Intelligent Networked Systems (INSI) in Boston, Massachusetts. During the trials, the setup demonstrated sustained throughput across multiple user devices under mobility conditions. Notably, the system managed beam control entirely at Layer 2.
Tommaso Melodia, director of INSI, stated that openness and performance do not require trade-offs in wireless design.
“You can have a fully open, reproducible system and still push the boundaries of what massive MIMO can deliver through algorithmic control,”
Tommaso Melodia, Director of INSI
The research team designed the open-source AI-RAN platform to be reproducible by other academic and industry researchers. Previously, the INSI team certified the AmpliTech radio unit for O-RAN conformance at the same Boston facility. As a result, this certification laid the groundwork for the interoperability achieved in the demonstration.
Industry Implications and Future Outlook
Industry analysts noted that using GPUs for network acceleration is a growing trend in the telecommunications sector. For example, Softbank and Nvidia previously demonstrated a GPU-based 16-layer massive MIMO system in an outdoor environment. However, that previous demonstration did not use open-source software.
Roy Chua, principal analyst at AvidThink, stated that the demonstration shows progress in using GPUs for network acceleration. Moreover, it shows how hardware providers are expanding the use of GPU architectures within the mobile telecom sector. This research could accelerate the adoption of artificial intelligence in next-generation networks.
