OAIC Framework

5G wireless is known for a revolution in flexibility in wireless networks. Beyond 5G and 6G will likely be known for using intelligence to capitalize on flexibility. Wireless research activities have made use of artificial intelligence (AI) in various areas, but practical implementations of AI-based algorithms in wireless networks are relatively rare.  Our goal is to provide a community research infrastructure that will allow you to experimentally verify your AI-enhanced wireless research and contribute to designing, prototyping, and testing 6G cellular communications features. Open AI Cellular lays out the design and architectural foundations of a software infrastructure that would spur experimental research, development and education on 6G wireless networks.

The OAIC framework leverages open source 5G software and implements the O-RAN interfaces and radio access network intelligent controllers (RICs) for enabling real-time experimentation with software radios in testbeds and production-like operating environments.

OAIC-C: AI Controllers

Open AI Cellular, or OAIC, is composed of AI-enabled radio access network controllers (OAIC-C) and a co-developed testing system (OAIC-T). More specifically, OAIC will be an AI-enabled cellular network created from open-source 5G software and re-engineered to have AI controllable parameters and AI testing software that will probe the network to find implementation, system-level, and security flaws in the AI-enabled radio access network.

Beyond xApps

  • Real-time control loop for AI-enhanced PHY layer control

  • Orchestration & Management (O&M) and rApps

OAIC-T: AI Testing Framework

While AI models are enablers to achieve intelligent wireless networks, the inability of the current theory to explain or prevent failures in the AI models emphasizes on the necessity of having a framework and appropriate environment for testing AI models deployed as cellular RAN controllers. There is a strong and urgent need for methods, platforms and tools that facilitate testing various AI models in the radio network in a production like environment. The AI testing framework (OAIC-T) that we propose will help to optimize the evaluation of AI algorithms under variations of the input data and learning models to maintain stable performance under realistic and, possibly, changing conditions. OAIC-T will also act as a support theme to define and incorporate system requirements, metadata, functional conditions, learning ability, and its collaborations with other agents. Given the large input space of an AI-controlled system, it might be impossible to exhaustively and comprehensively test performance of AI models without the help of AI. OAIC-T provides a scalable and open-source AI-enabled testing platform for AI-controlled radio access networks, enabling automated and iterative testing.

Testbeds

While advanced wireless testbeds for 5G research are being deployed in the United States, we propose to develop a community research infrastructure that will unleash 6G research and experimentation platforms. The project will supplement open software-defined radio (SDR) architectures with artificial intelligence (AI) radio access network controllers. More precisely, our software will enhance open-source 5G efforts to provide a functional and scalable software-defined platform that implements and advances the O-RAN specifications. It will provide a framework for prototyping and testing of AI-driven 6G cellular radio access networks and exploring new features, applications, and use cases. You will be able to download and modify the source code for free and readily use the software for diverse wireless experiments on major wireless community research testbeds. 

Laboratory Testbeds

  • Small-scale 5G O-RAN laboratory testbeds

  • Features near-real time RIC, 5G NSA and SA, developed using srsRAN and O-RAN software, and 2-14 SDRs and ZeroMQ simulator integrated with GNU Radio

Virginia Tech CORNET

  • Remotely accessible in-building  testbed with numerous USRPs, switches, and virtual machines

  • Develop source code, automate testing, deploy applications

  • Users can run sample OAIC experiments or develop and execute new experiments on CORNET

NSF PAWR Platforms

  • City-scale outdoor testbeds featuring USRPs and other radios, drones (AERPAW), networking and computing resources, FCC licenses

  • In collaboration with PAWR PIs, OAIC will be installed and tested on PAWR platforms and made available to their users