AI/ML Model Lifecycle Management for UE Capability Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current telecommunication standards lack specification support for ensuring AI/ML model inference complexity within UE capability and do not enable effective performance monitoring and management of AI/ML functionalities in new radio (NR) systems, particularly for channel state information (CSI) compression and beam prediction.
Innovation Solution
Implement functionality-based and model ID-based life cycle management (LCM) systems and methods for UEs and gNBs to manage AI/ML models based on UE capability reports, including reporting properties and maximum numbers of supported functionalities and active models, and configuring functionalities to comply with UE capabilities, with mechanisms for data collection and performance monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If AI/ML functionalities are introduced in NR systems, then system-level benefits and complex function inference capabilities are improved, but inference complexity exceeds UE capability and device complexity increases
Solution Approach 1:
The patent segments AI/ML functionalities into distinct use cases (CSI compression, CSI prediction, beam prediction) with individual capability reporting for each. This allows the UE to selectively support specific AI/ML functionalities based on its processing capabilities, rather than requiring support for all possible functionalities simultaneously.
Solution Approach 2:
The patent introduces capability parameters (maxNumberOfAImlActiveFunctionalities, properties of supported functionalities) that allow the UE to dynamically indicate its processing capacity. The gNB can then adjust the configuration of AI/ML functionalities based on these parameters, changing the operational parameters to match UE capabilities.
2Measurement precision
If performance monitoring is implemented at UE side, then model accuracy and functionality performance are improved, but measurement and monitoring complexity increases
Solution Approach 1:
The patent implements a feedback mechanism where the UE monitors the performance of AI/ML models (accuracy, latency) and reports this information back to the gNB. The gNB uses this feedback to make informed decisions about model updates, retraining, or configuration adjustments, creating a closed-loop system that improves performance while distributing monitoring responsibilities.
Solution Approach 2:
The patent introduces performance monitoring entities and reference models as intermediaries. Instead of requiring the UE to perform all monitoring functions directly, reference models and monitoring entities act as mediators that simplify the measurement process while maintaining accuracy, reducing the burden on the UE's processing capabilities.
3Speed
If dedicated processing units are used for AI/ML operations, then inference speed and processing capability are improved, but device complexity and resource requirements increase
Solution Approach 1:
The patent enables dynamic selection and configuration of processing resources for AI/ML operations. The UE can adaptively use dedicated processing units (such as neural processing units) when available and needed, while falling back to general-purpose processors when dedicated units are not available or when the complexity of the AI/ML task does not justify their use. This dynamic approach optimizes speed while managing device complexity.
Data Source
AI summary
A system and a method are disclosed for AI/ML model LCM. A method performed by a UE includes transmitting, to a base station, a plurality of properties of supported functionalities for each of a plurality of AI/ML use cases, wherein each of the plurality of AI/ML use cases is configured with an individual use case ID; transmitting, to the base station, a maximum number of AI/ML active functionalities that the UE supports across the plurality of AI/ML use cases; receiving, from the based station, a report configuration based on the plurality of properties of supported functionalities for each of the plurality of AI/ML use cases and the maximum number of AI/ML active functionalities that the UE supports across the plurality of AI/ML use cases; generating a report based on the report configuration; and transmitting the report to the base station.