UE Measurement Reporting for Unbiased AI Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in providing unbiased training data for network-side AI/ML models due to event-triggered measurement reporting, which leads to biased training data and resource wastage, particularly when UEs are non-mobile.
Innovation Solution
Implementing a configuration mechanism for measurement and reporting objects that activates subsets based on triggers, such as MAC-CE, higher layer signaling, or probabilistic triggers, to collect and report measurements only when specific conditions are met, ensuring an equal distribution of measurements over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If event-triggered measurement reporting is used, then measurement reporting is simplified, but training data becomes biased and resources are wasted
Solution Approach 1:
The patent implements periodic measurement reporting where UEs report measurements at regular intervals regardless of event triggers. This periodic approach ensures unbiased training data collection while reducing resource wastage by establishing predictable reporting patterns that avoid redundant event-triggered transmissions.
2Device complexity
If event-triggered measurement reporting is used, then reporting complexity is reduced, but training data bias increases
Solution Approach 1:
The patent employs periodic reporting intervals that ensure measurements are collected systematically over time, providing unbiased training data for AI/ML models. This periodic approach maintains manageable reporting complexity while significantly improving training data quality by eliminating the bias inherent in event-triggered reporting.
Solution Approach 2:
The patent changes the reporting parameter from event-based triggering to time-based periodic intervals. This parameter change fundamentally improves training data quality by ensuring representative sampling of channel conditions, while the periodic nature keeps the implementation complexity manageable through standardized timing mechanisms.
3Loss of information
If all measurement objects are activated continuously, then complete data collection is achieved, but resource usage increases
Solution Approach 1:
The patent segments measurement objects into different groups with different activation patterns. Instead of continuously activating all measurement objects, the system divides them into subsets that are activated periodically or based on specific criteria, thereby maintaining complete data collection capability while reducing overall resource usage through selective activation.
Solution Approach 2:
The patent applies periodic activation to measurement objects rather than continuous activation. Measurements are collected at regular intervals across different measurement objects, ensuring complete data collection over time while reducing instantaneous resource usage by activating only the necessary subset at each period.
Data Source
AI summary
Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a user equipment (UE) may receive a configuration that indicates one or more measurement objects and one or more reporting or logging objects, wherein a first subset of the one or more measurement objects and the one or more reporting or logging objects are activated at the configuration and a second subset of the one or more measurement objects and the one or more reporting or logging objects are deactivated at the configuration. The UE may transmit, based at least in part on a trigger that activates the one or more measurement objects or the one or more reporting or logging objects, a measurement report that indicates training data based at least in part on the one or more measurement objects or the one or more reporting or logging objects. Numerous other aspects are described.