ML Positioning Model Monitoring via Dual Reference Signal Configuration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in accurately monitoring and managing the life cycle of machine learning (ML) positioning models, particularly due to the sensitivity of ML models to environmental changes and the lack of efficient methods for establishing a 'ground truth' for monitoring metrics.
Innovation Solution
The system configures a user equipment (UE) to receive configuration messages for two sets of reference signals, allowing it to perform measurements and generate data for both positioning and monitoring purposes. The UE then transmits reporting messages based on these measurements, enabling network entities to monitor the accuracy of ML positioning models and perform life cycle management operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ML positioning models are used to improve positioning accuracy, then positioning precision is improved, but the models become sensitive to environmental changes and UE mobility requiring frequent monitoring
Solution Approach 1:
The patent implements a feedback mechanism where monitoring measurements are continuously collected from UEs and fed back to the network entity. This feedback loop enables the system to detect when ML positioning model accuracy degrades due to environmental changes or UE mobility, triggering appropriate responses such as model retraining or switching to alternative positioning methods, thus maintaining reliability while preserving positioning accuracy.
Solution Approach 2:
The patent performs preliminary actions by establishing ground truth measurements using high-accuracy positioning techniques (GPS, GNSS, Lidar) before deploying ML positioning models. These preliminary ground truth measurements are stored and used as reference data for subsequent monitoring and validation of ML model performance, enabling proactive detection of model degradation before it affects positioning accuracy.
2Measurement precision
If ground truth measurements are obtained from GPS, GNSS, or Lidar to establish accurate monitoring metrics, then measurement precision is improved, but power consumption and resource usage increase
Solution Approach 1:
The patent applies partial action by using ground truth measurements from power-intensive techniques (GPS, GNSS, Lidar) only when necessary for establishing initial monitoring metrics and validating ML model performance, rather than continuously. For ongoing monitoring, the system uses lower-power ML-based measurements compared against the stored ground truth, reducing power consumption while maintaining measurement precision where critical.
Solution Approach 2:
The patent performs preliminary action by obtaining ground truth measurements once during initialization or model validation phases, storing these measurements for future reference. This eliminates the need for continuous use of power-intensive positioning techniques, as the pre-obtained ground truth serves as a reference for comparing against ongoing lower-power ML positioning measurements.
3Reliability
If frequent monitoring of ML positioning models is performed to ensure accuracy, then positioning reliability is improved, but communication overhead increases
Solution Approach 1:
The patent applies local quality by implementing differentiated monitoring strategies based on local conditions. The system adjusts monitoring frequency and intensity according to UE mobility state, environmental stability, and ML model performance metrics. In stable environments with stationary UEs, monitoring is reduced to minimize overhead, while in dynamic conditions, monitoring intensity increases locally to maintain reliability, rather than applying uniform frequent monitoring network-wide.
Solution Approach 2:
The patent implements dynamic monitoring where the monitoring frequency and resources are adjusted in real-time based on observed conditions. The system dynamically scales monitoring intensity according to ML model performance degradation rates, UE mobility patterns, and environmental changes, optimizing the balance between positioning reliability and communication overhead through adaptive resource allocation.
4Measurement precision
If two sets of reference signals are configured for positioning and monitoring, then positioning accuracy and monitoring capability are improved, but device complexity increases
Solution Approach 1:
The patent applies universality by designing reference signals that serve multiple functions simultaneously. The same reference signal infrastructure is used for both positioning measurements and ML model monitoring, with the network entity configuring reference signals that can be measured for both purposes. This multi-functional approach reduces device complexity compared to implementing separate dedicated signal configurations for positioning and monitoring.
Solution Approach 2:
The patent merges positioning and monitoring functions by combining them into a unified measurement framework. Instead of implementing separate signal configurations and processing pipelines for positioning and monitoring, the system combines these functions to share common infrastructure, reducing device complexity while maintaining the accuracy benefits of dedicated measurements for both purposes.
Data Source
AI summary
This disclosure provides systems, methods, and devices for wireless communication that support machine learning (ML) positioning model monitoring and life cycle management. In some aspects, a user equipment (UE) may receive, from a network entity, a configuration message associated with a first set of reference signals and a second set of reference signals. The UE may perform first measurements based on the first set of reference signals received from a first set of transmit/receive points (TRPs) to generate first measurement data. The UE may perform second measurements based on the second set of reference signals received from a second set of TRPs to generate second measurement data. The UE may transmit, to the network entity, a reporting message based on the second measurements or based on the first measurement data and the second measurement data. Other aspects and features are also claimed and described.