RanControl

Next-generation networks must handle highly diverse application requirements—from industrial automation, connected vehicles to 4K video streaming. Open Radio Access Networks (ORAN) are at the forefront of this shift, allowing standardised data-driven control of the Radio Access Network (RAN).

However, building ML-driven intelligent closed-loop control systems that can generalise to new tasks remains a core challenge. RanControl developed reinforced continual graph learning xApps that adapt to changing network tasks without catastrophic forgetting. The approach leverages structural reinforcement learning and few-shot learning to enable task-incremental model training, marking a new step in the evolution of intelligent RAN automation.

The key objectives of RanControl are:

1. Developing robust models for RAN scheduling and slicing.
2.Deploying ML models as containerised xApps within virtualised RAN stacks.
3.Conducting large-scale performance evaluation across diverse network scenarios to ensure scalability, adaptability, and industry readiness.