Optimizing UE Power Efficiency: AI/ML Approach for Upgrade Time Determination

Abstract

The transition of Control Plane (CP) block functions into software entities, as proposed by 3GPP, necessitates periodic downtime for maintenance activities such as software upgrades or failures. This downtime requires the disconnection of all User Equipment (UE) connections to the CP, triggering the UE reattach procedure and resulting in increased UE power consumption and spectrum wastage. To mitigate these challenges, optimal CP upgrade timings should align with periods of low traffic. In this paper, we propose an AI/ML-based procedure to autonomously determine the optimal time to upgrade CP block functions, eliminating the need for manual intervention by operators. Our approach involves analyzing traffic conditions using statistical data from several CP blocks managing base stations across various areas, including residential and non-residential zones like subways, shopping complex and hospitals. Leveraging Seasonal Auto-Regressive Integrated Moving Average (SARIMA) forecasting, we predict bearer statistical data to calculate the optimal CP software upgrade time, validated using Z-Score analysis at the same time. In addition to address the suboptimal upgrade timings, we also proposed CP Outage Handling Procedure (COHP) v2 by preserving UE contexts during CP upgrades. Our results demonstrate SARIMA’s high accuracy in predicting lean traffic conditions, with an R-Squared score of 0.99. Furthermore, upgrading CP software during predicted lean periods leads to substantial UE power savings ranging from 80% to 97% compared to manual upgrades.

Keywords

• 5G core
• cloud native
• control plane
• user plane
• CNF
• AI/ML