ML Positioning Model Mitigating UE Clock Drift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning (ML)-based positioning in wireless communication systems is sensitive to user equipment (UE) clock drift, which can lead to inaccuracies in predicted positions due to timing variations caused by clock drift.
Innovation Solution
A method where a user equipment (UE) receives positioning configuration information from a network entity, performs positioning operations, and transmits positioning measurements to a training entity to enable the training of an ML positioning model that accounts for UE clock drift.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ML positioning model is trained with timing information from UEs, then positioning accuracy is improved, but UE clock drift causes timing variations that degrade positioning accuracy
Solution Approach 1:
The system performs preliminary actions by collecting multiple positioning measurements from multiple UEs before training the ML model. This preliminary data collection phase allows the model to learn from aggregated timing information that inherently averages out individual UE clock drift effects, thereby improving positioning accuracy despite unreliable timing from individual UEs
Solution Approach 2:
The ML positioning model acts as an intermediary that processes timing information from multiple UEs. By training the model with diverse timing data from multiple sources, the system creates a robust intermediary that can compensate for clock drift in any single UE, thus maintaining reliable positioning accuracy
2Productivity
If ML positioning model uses timing variations to predict position, then positioning operations are performed, but clock drift translates small timing variations into large position errors
Solution Approach 1:
The system performs preliminary training of the ML model using timing data from multiple UEs before actual positioning operations. This preliminary training phase enables the model to learn the relationship between timing variations and positions while accounting for clock drift, so that during productivity-focused positioning operations, accurate predictions can be made without real-time clock synchronization
Solution Approach 2:
The system collects more timing measurements than strictly necessary for basic positioning by utilizing multiple UEs. This excessive data collection approach provides redundant information that helps compensate for clock drift, allowing the model to achieve accurate position predictions even when individual timing measurements are affected by drift
Data Source
AI summary
This disclosure provides systems, methods, and devices for wireless communication that support machine learning (ML)-based positioning that mitigates user equipment (UE) clock drift. In some aspects, a UE may receive, from a network entity, positioning configuration that indicates positioning operations to be performed to gather training data to train an ML positioning model to account for UE clock drift. The UE may monitor for positioning reference signals from a transmit/receive point and transmit positioning measurements to a training entity. The positioning measurements may include multiple measurements at a fixed location, or the UE may augment the positioning measurements based on simulated clock drift measurements. Alternatively, the UE may transmit clock drift information with the positioning measurements to the training entity. Alternatively, the UE may utilize a hybrid approach that combines multiple positioning measurements with augmentation or providing clock drift information. Other aspects and features are also claimed and described.