/
5G RAN Slice PRB Prediction

5G RAN Slice PRB Prediction

Code / Gerrit:

5G RAN Slice PRB Prediction Rapp:

gerrit.o-ran-sc Code Review - nonrtric/plt/rappmanager.git/tree - sample-rapp-generator/rapp-slice-prb-prediction/

5G RAN NSSMF Simulator:

gerrit.o-ran-sc Code Review - nonrtric/plt/rappmanager.git/tree - sample-rapp-generator/rapp-ran-nssmf-simulator/

 

Slides:

https://lf-onap.atlassian.net/wiki/download/attachments/428179457/2025-09-26_5G_RAN_Slice_PRB_Prediction_Rapp_Demo.pdf

Demo:

https://zoom.us/rec/share/cLSXyy1yALrZUJPGMKD8vZBCz8U1ql__3Gkoe1uiw4zRI7p7nfuqF_9wHfS7BBgN.OR8bD5EC1P_mQ484

Goal:

  • Implement 5G RAN Slice PRB Prediction Use Case using O-RAN SC Non-RT RIC, RAN NSSMF Simulator and 5G RAN Slice PRB Prediction Rapp.

  • RAN resources specially PRBs are finite and shared across slices. Without proper PRB allocation can lead to packet drops, latency spikes or QoS violations.

  • Multi-slice resource optimization – without prediction, slice scheduling relies on static or historical data. In dynamic networks, this leads to over-provisioning some slices, while starving others.

  • To solve this, we adopted AI/ML algorithms to predict future PRB usage in 5G RAN slice using historical data. Predicted data is used in multi-slice resource optimization.

Steps:

Future Work:

Use O-RAN SC AI/ML Framework AI/ML Framework - AI/ML Framework - Confluence/Wiki to implement this use case, so 5G RAN Slice PRB Prediction Rapp becomes lighter. As of now Rapp contains offline trained AI/ML models to predict future 5G RAN Slice PRB using historical data.