5G Session Continuity via Predictive PSA UPF Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current reselecting schemes for packet data unit (PDU) session anchors (PSA) UPFs in 5G core networks result in significant signaling traffic and negatively impact service quality, particularly for high-speed user equipment like autonomous vehicles and drones, due to frequent changes in service areas.
Innovation Solution
A method and device that utilize machine learning to predict the residence time of user equipment in service areas, allowing for intelligent selection of PSA UPFs based on mobility patterns, reducing unnecessary reselections and signaling traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current reselecting schemes are used for PSA UPF, then service area coverage is maintained, but signaling traffic increases significantly and service quality deteriorates
Solution Approach 1:
The patent applies preliminary action by predicting the UE's future service area and residence time before the actual movement occurs. The prediction unit uses machine learning models to forecast which UPF the UE will likely connect to next, allowing the network to prepare and execute PSA UPF reselection proactively rather than reactively, thereby reducing unnecessary signaling traffic while maintaining service quality
Solution Approach 2:
The patent changes the parameter of decision-making from immediate location-based triggers to prediction-based residence time estimates. By introducing predicted residence time as a key parameter, the system determines whether PSA UPF reselection should occur, transforming the reselection criterion from purely location-dependent to a combination of location and time-prediction factors, thus reducing signaling traffic
2Speed
If frequent PSA UPF reselection is performed, then service area tracking is improved, but service quality for high-speed UE deteriorates
Solution Approach 1:
The prediction unit performs preliminary analysis of UE mobility patterns using machine learning to forecast future service areas and residence times. This allows the system to determine in advance whether PSA UPF reselection is necessary, preventing premature or unnecessary reselection events that would degrade service quality for high-speed UE while maintaining adequate tracking capability
Solution Approach 2:
The system implements feedback by continuously monitoring actual UE movement patterns and comparing them with predictions. The prediction model is refined using historical mobility data, allowing the system to adapt to UE behavior patterns and improve the accuracy of residence time predictions, thereby optimizing the balance between tracking speed and service quality
3Quantity of substance
If PSA UPF reselection is optimized using prediction, then signaling traffic is reduced, but device complexity increases
Solution Approach 1:
The patent introduces a prediction unit as an intermediary component between the mobility management function and the PSA UPF reselection mechanism. This intermediary uses machine learning models to process mobility information and generate residence time predictions, adding intelligence to the decision-making process while keeping the core network functions relatively simple and modular
Solution Approach 2:
The prediction unit operates autonomously, self-managing the collection of mobility information, training of prediction models, and generation of residence time estimates without requiring complex external control systems. The system uses available mobility data to automatically refine its predictions, reducing the need for additional complex infrastructure
Data Source
AI summary
In embodiments, a session management device includes at least one transceiver, and at least one processor coupled to the at least one transceiver. The at least one processor is configured to obtain prediction information for a residence time of a user equipment (UE) per a service area of each user plane function (UPF), based on mobility information of the UE. The at least one processor is configured to identify a packet data unit (PDU) session anchor (PSA) UPF for the UE based on the prediction information. The at least one processor is configured to, based on identifying an event according to a movement of the UE, identify whether a UPF corresponding to the service area in which the UE is located is the PSA UPF for the UE. The at least one processor is configured to, based on identifying that the UPF is the PSA UPF for the UE, perform a PSA change to the UPF.