5G AI Model Signaling for Consistent UE and Network Handover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3GPP 5G technology lacks methods and apparatuses for effective AI model management, particularly in scenarios involving UE transitions between RRC states, model updates, and handovers, leading to inconsistencies in AI model availability and management across network nodes.
Innovation Solution
Implement signaling procedures for AI model management, including instructions for model handling, model ID and delta part exchange, and UE capability-based decisions to ensure consistent AI model availability and management across network nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI model management methods are implemented in 3GPP 5G networks, then AI model availability and consistency across network nodes is improved, but system complexity increases due to additional signaling procedures
Solution Approach 1:
The AI model management signaling is segmented into distinct message types: AI model information messages carry model definitions and parameters, while AI model management messages handle operations like updates and deletions. This segmentation allows each message type to be optimized for its specific function, improving reliability without overwhelming the system with monolithic complex procedures.
Solution Approach 2:
The network node receives AI model information from the UE before initiating management operations. This preliminary action ensures that the network has knowledge of available models, their capabilities, and constraints prior to performing updates or deletions, thereby ensuring consistent AI model availability across network nodes without requiring complex real-time negotiation.
2Productivity
If AI model updates are performed based on default models, then model efficiency is improved by reducing redundant data, but information completeness may be compromised
Solution Approach 1:
The patent extracts only the differential portions (delta parts) of AI models from the complete model definitions. When an AI model is updated based on a default model, only the changed parameters and structures are transmitted and stored. This extraction approach maintains information completeness for the specific update context while dramatically improving update efficiency by eliminating redundant unchanged data.
Solution Approach 2:
Instead of transmitting complete AI model definitions for every update, the system performs partial action by sending only the necessary delta parts. The network node and UE maintain the complete model context by combining received delta parts with stored default models, achieving information completeness through partial information exchange that optimizes productivity.
3Adaptability or versatility
If UE capability-based decisions are used for AI model handling, then adaptability is improved, but processing time increases
Solution Approach 1:
The UE provides capability information to the network node in advance, before AI model management operations are initiated. This preliminary action includes memory size constraints, supported model types, and other relevant capabilities. The network node uses this pre-provided capability information to make immediate adaptability-based decisions without requiring time-consuming queries during model handling operations.
Solution Approach 2:
The system implements feedback through the capability information exchange mechanism. The UE's capability report serves as feedback to the network node, enabling the network to adapt its AI model management strategy to the specific UE's constraints and capabilities. This feedback loop establishes adaptability while maintaining efficiency by avoiding repeated capability assessments during model operations.
Data Source
AI summary
Embodiments of the present application relate to methods and apparatuses for artificial intelligence (AI) model management. According to an embodiment of the present application, a network node includes a processor and a transceiver coupled to the processor; and the processor is configured to: receive, via the transceiver, information related to a set of artificial intelligence (AI) models available at a user equipment (UE); and transmit, via the transceiver, information indicating at least one of the following: information of an AI model that is not included in the set of AI models; or handling the set of AI models.