SoftBank

Welcome to Transformer-based Inter-Cell Coordination on OCUDU

Advancing AI-RAN through Multi-Cell Intelligence and Real-Time Coordination

What is this project?

This project is a collaboration between SoftBank and CCI/Virginia Tech focused on extending SoftBank’s AITRAS framework from single-cell AI optimization toward intelligent multi-cell coordination in 5G radio access networks. The project aims to develop a transformer-based architecture that can jointly optimize advanced RAN functions such as inter-cell coordination, multi-site scheduling, MU-MIMO, user pairing, and interference management.

By integrating this capability with the open-source OCUDU CU/DU platform, the project seeks to create a practical AI-RAN deployment pathway where machine learning models can operate directly within real-time RAN infrastructure. The broader goal is to transform AITRAS from a single-cell optimization framework into a scalable multi-cell coordination system for dense and high-performance 5G networks.

Project Objectives

The primary mission of this project is to enable real-time, AI-driven coordination across multiple 5G cells while maintaining the latency requirements needed for practical RAN deployment. Key objectives include:

Extending AITRAS to Multi-Cell Coordination:
Developing transformer-based models that can jointly process multi-cell channel, user, and interference information to support coordinated scheduling and resource allocation across neighboring cells.

Enabling Advanced MU-MIMO Capabilities:
Supporting high-order MU-MIMO through intelligent user pairing, spatial resource allocation, and transformer-based multi-user detection to improve cell capacity and spectral efficiency.

Supporting Real-Time RAN Adaptability:
Designing AI models that can adapt to dynamic traffic patterns, user mobility, and changing channel conditions within sub-millisecond latency constraints.

Integrating AI-RAN with OCUDU:
Bringing AITRAS capabilities into the OCUDU platform to enable real-time AI model deployment inside an open CU/DU architecture.

Advancing Commercial AI-RAN Deployment:
Creating a pathway for transformer-based RAN intelligence to move beyond research prototypes and toward deployment-ready, standards-aligned 5G infrastructure.

Project Impact

This project addresses one of the key challenges in next-generation wireless networks: moving from isolated cell-level optimization to coordinated, intelligent, multi-cell operation. By combining SoftBank’s AITRAS framework with OCUDU, the project aims to demonstrate how AI can improve RAN capacity, reduce interference, and enable more adaptive 5G deployments in dense urban, enterprise, and future AI-RAN environments.