Wireless Positioning Training Data Collection for Non-LOS Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
LTE/NR positioning methods may not perform well in conditions without line of sight (LOS) measurements, and existing AI/ML-based solutions may not adequately address this issue.
Innovation Solution
A Model Training logical function (MTLF) is employed to train AI/ML models by determining target areas and collecting positioning training data from PRUs or RAN nodes, considering environmental and mobility characteristics, and sending collection requests to LMFs for data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If geometry-based positioning methods are used, then the system is simple to implement, but positioning accuracy deteriorates when line of sight measurements are not available
Solution Approach 1:
The patent replaces traditional geometry-based positioning methods with AI/ML-based algorithms. The network device collects positioning measurement data from multiple cells and uses trained machine learning models to determine UE location, substituting the mechanical/geometric calculation approach with an intelligent system that can handle non-line-of-sight conditions effectively.
Solution Approach 2:
The patent changes the input parameters for positioning by collecting multiple types of positioning measurement data (such as RSRP, RTOA, R-TDoA) from multiple cells instead of relying on traditional geometric parameters. The AI/ML models process these varied parameters to achieve accurate positioning in complex environmental conditions.
2Measurement precision
If AI/ML-based positioning algorithms are implemented, then positioning accuracy improves in non-LOS conditions, but system complexity increases
Solution Approach 1:
The patent introduces a Model Training Logical Function (MTLF) and Location Management Function (LMF) as intermediaries. The MTLF trains the AI/ML models using collected positioning data, and the LMF executes the trained models during actual positioning operations. This intermediary architecture separates the complex model training process from the real-time positioning function, managing system complexity effectively.
Solution Approach 2:
The patent performs preliminary action by training the AI/ML models in advance using the MTLF before actual positioning operations. The model training phase collects positioning measurement data and trains the models offline, so that during real-time positioning, only the pre-trained models need to be executed, reducing the complexity of real-time processing.
3Reliability
If targeted training data collection is performed, then model training effectiveness improves, but data collection complexity increases
Solution Approach 1:
The patent applies local quality by determining specific target areas for data collection based on the characteristics of the positioning model being trained. The MTLF identifies geographic areas and environmental conditions that are most relevant for training, and directs data collection efforts to those specific locations rather than collecting data uniformly across all areas, thus improving training effectiveness while managing collection complexity.
Data Source
AI summary
A network device may receive a request for positioning training data that includes an indication of environmental characteristics associated with the positioning training data, and determine a target area to collect the positioning training data based on the request. The target area may be based on the environmental characteristics associated with the positioning training data or availability of one or more PRUs or RAN nodes that are capable of providing the positioning training data. The network device may determine a list of PRUs that can provide the positioning training data, determine a positioning measurement type associated with the positioning training data, and send a collection request for the positioning training data, where the collection request may include an indication of the target area, the list of the PRUs, and the positioning measurement type. The network device may receive a positioning training data report that includes the positioning training data.