PAGE CONTENTS
Objectives
3F.047 aims to enhance radio resource management in densely populated areas through machine learning applied to open-source 5G-Non-Terrestrial Network (NTN) systems. The objective is to produce a software solution for high performance and efficient resource allocation, for resources such as channels, modulations, and beams, within low size, weight, and power (SWaP) base stations. This project will produce technology that enables robust, secure, and high-capacity connections in previously unreachable locations and will be demonstrated on a prototype hardware system using commercial-off-the-shelf components. The project sets the stage for high-performance regenerative gNodeB’s as part of future NTN networks.
Challenges
The key challenges of the project involve building an inference subsystem that can handle large volumes of IQ data with very low latency in a very resource constrained package.
System Architecture
The system gNodeB consists of a Commercial Off-The-Shelf (COTS) Universal Software Radio Peripheral (USRP) on a low size, weight and power processor (ARM or x86-based) which runs all software to run a RAN, including a modified version of srsRAN, an open5GS 5G core, and the Qoherent Radio Inference Engine.
Plan
The project is split into 6 milestones:
- Requirements Analysis, Alignment and Design
- RIARAN-NTN GnodeB & Network Development
- ML Workflow & Model Development
- Integration and Lab Testing
- Field Testing and Validation
- Project reporting and closure
Current Status
Qoherent has done preliminary design work, hardware target benchmarking and performed field testing to inform testbed requirements. Development activities within the project such as development of the scheduler and construction of the project testbed is underway.
Related Links
Companies
