UE Network Slice Reselection via ML Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In wireless communication systems, particularly in 5G NR, there is a need to improve communication performance by predicting the next active PDU session and switching to the corresponding frequency channel efficiently, as existing methods lead to delays and service disruptions during connection setup, especially when multiple PDU sessions with different network slice instances are involved.
Innovation Solution
A method where user equipment (UE) predicts the next active PDU session based on past connection setup events and selects the appropriate frequency channel to camp on during idle or inactive modes, using network slice information to prioritize frequency channels associated with each PDU session, thereby reducing delays and service disruptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If the UE camps on a frequency channel based on traditional methods without prediction, then the frequency channel selection is simple, but connection setup delays occur and service disruptions happen when switching between multiple PDU sessions
Solution Approach 1:
The UE performs prediction of the next active PDU session in advance while in idle or inactive mode, determining the frequency channel to camp on before actually needed. This preliminary action allows the UE to switch frequency channels proactively rather than reactively, eliminating connection setup delays when transitioning between PDU sessions associated with different network slice instances
Solution Approach 2:
A machine learning model acts as an intermediary between past connection setup events and future frequency channel selection. The model processes historical data including inter-connection setup times and PDU session patterns to predict which PDU session will become active next, thereby mediating the complex decision-making process and simplifying the overall system while improving performance
2Productivity
If the UE uses machine learning to predict the next active PDU session, then connection setup efficiency improves, but the processing complexity and computational requirements increase
Solution Approach 1:
The machine learning model is deployed locally on the UE itself, enabling self-service prediction without requiring external network assistance or complex centralized processing. The UE autonomously collects its own connection setup events, trains and updates the model with its own historical data, and performs predictions independently, thereby improving connection setup efficiency while keeping the overall system architecture relatively simple
Solution Approach 2:
Instead of implementing complex real-time analysis of all possible PDU session scenarios, the system uses a simplified machine learning model that copies essential patterns from historical data. The model captures the temporal relationships and decision-making logic from past connection setup events, creating a lightweight predictive mechanism that achieves high efficiency without replicating the full complexity of network slice management
Data Source
AI summary
Apparatus, methods, and computer-readable media for facilitating network slice based reselection using machine learning are disclosed herein. An example method for wireless communication at a UE includes predicting a next active PDU session of two or more PDU sessions based on next connection setup times for each PDU session. The next connection setup times may be estimated based on connection setup events occurring over a period. The example method also includes selecting a frequency channel to camp on based on the next active PDU session.