AI Inference Models for Wireless Positioning Measurement Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wireless communication systems face challenges in efficiently measuring and reporting positioning data for artificial intelligence (AI) based positioning, particularly in distinguishing between line-of-sight (LOS) and non-line-of-sight (NLOS) paths.
Innovation Solution
The system enhances measurement and reporting by enabling a request indication from a location server for labelled or unlabelled positioning measurements, depending on whether a supervised or unsupervised learning model is used. This allows for the training and deployment of AI inference models at the location server, UE, or base station, enabling improved classification and prediction of positioning metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional positioning measurement methods are used, then the system can obtain basic positioning data, but the accuracy in distinguishing LOS and NLOS paths is insufficient
Solution Approach 1:
The patent introduces AI inference models as intermediaries between the raw positioning measurements and the final positioning results. These models process the measurement data and provide enhanced classification of LOS and NLOS paths, thereby improving both measurement precision and reliability simultaneously
Solution Approach 2:
The patent transforms the positioning measurement data by applying AI-based parameter transformations. The inference models learn optimal parameter representations from training data and use these transformed parameters to achieve more accurate and reliable path distinction than traditional methods
2Measurement precision
If AI inference models are trained and deployed at the location server, then positioning prediction accuracy improves, but system complexity and training resource requirements increase
Solution Approach 1:
The patent segments the AI model deployment architecture into multiple options: location server-based deployment, UE-based deployment, or distributed deployment. This segmentation allows the system to choose the appropriate complexity level based on specific requirements, balancing accuracy improvements against system complexity
Solution Approach 2:
The patent implements preliminary training of AI inference models using historical positioning data before actual positioning operations. This preliminary action allows the model to learn from extensive training data in advance, achieving high prediction accuracy while keeping the operational system relatively simple
3Reliability
If labelled positioning measurements are collected and processed, then AI model training quality improves, but measurement and processing time increases
Solution Approach 1:
The patent performs labelling of positioning measurements in advance during the model training phase. By preparing labelled training data beforehand, the system achieves high training quality without adding time delays to the actual positioning operations, as the labelling work is completed during offline training
Solution Approach 2:
The patent applies partial labelling strategies where only critical measurements or a subset of measurements require detailed labelling. This partial action approach maintains sufficient training quality while reducing the overall time and computational resources needed for the labelling process
Data Source
AI summary
Various aspects of the present disclosure relate to a device that receives a request by a location server to query for positioning measurements which are labelled or unlabelled, which is dependent on whether a supervised or unsupervised learning model is used. The requested positioning measurements, each associated with a label, are provided to a training system that trains one or more artificial intelligence (AI) inference models. The one or more AI inference models are deployed to a location management function (LMF), which makes predictions based on subsequent positioning measurements received from the device.