UE Positioning Model Validation Using Reference Device Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wireless communication systems lack effective mechanisms for validating and testing machine learning models used for UE positioning, particularly in environments where the models may become inaccurate due to changes in UE position or environmental conditions, and network nodes have limited knowledge about UE-based models.
Innovation Solution
A method and apparatus for validating machine learning models used for UE positioning by receiving positioning measurement reports and reference positioning-related information from reference devices, determining performance indications, and taking actions based on these indications to ensure model accuracy and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If machine learning models are deployed for UE positioning without validation mechanisms, then positioning service responsiveness is improved, but positioning accuracy deteriorates due to model inaccuracy from environmental changes
Solution Approach 1:
The patent implements preliminary validation actions by collecting reference positioning measurements and ground truth data before deploying ML models for positioning. The system pre-establishes validation datasets from reference devices and performs model accuracy assessment in advance, ensuring models are validated against known reference data before actual positioning operations begin.
Solution Approach 2:
The patent establishes a feedback mechanism where ML model performance is continuously monitored by comparing model-generated positioning estimates against reference measurements from trusted devices. The system collects feedback on model accuracy and uses this information to determine when model updates or retraining are needed, creating a closed-loop validation system that maintains positioning accuracy over time.
2Measurement precision
If ML model validation mechanisms are implemented, then positioning accuracy is improved, but system complexity increases due to additional validation infrastructure
Solution Approach 1:
The patent implements self-service validation where the network infrastructure itself provides reference measurements and validation data without requiring external third-party systems. Network nodes and reference devices within the existing cellular infrastructure generate the validation datasets, allowing the system to validate its own ML models using internally-generated reference data rather than external complex validation infrastructure.
Solution Approach 2:
The patent makes existing network components multi-functional by having reference devices serve dual purposes: they perform normal positioning measurements for UEs while simultaneously providing reference data for ML model validation. This eliminates the need for separate dedicated validation infrastructure, as existing network elements are utilized for both service delivery and model validation functions.
3Measurement precision
If reference devices are used for model validation, then model accuracy assessment is improved, but network resource consumption increases
Solution Approach 1:
The patent implements partial validation action by selectively applying ML model validation only when necessary - such as during model updates, when positioning accuracy requirements change, or based on performance thresholds. Rather than continuously validating models at all times, the system performs validation at strategic intervals and conditions, reducing unnecessary network resource consumption while maintaining adequate model accuracy assessment.
Data Source
AI summary
A method may include receiving, by a first user device from a network node or a second user device, 1) a positioning measurement report including at least one positioning measurement measured by a reference device, and 2) reference positioning-related information to be used for testing and/or validating a machine learning model; determining estimated positioning-related information as outputs of the machine learning model based on at least a portion of the positioning measurement report as inputs to the machine learning model; determining a performance indication of the machine learning model based on the reference positioning-related information and the estimated positioning-related information, wherein the performance indication indicates a performance or accuracy of the machine learning model; and performing, by the first user device, an action based on the performance indication.