Label Information Correction for NLOS Positioning Model Training

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing AI/ML models for terminal device positioning in communication systems face challenges in achieving high accuracy due to unreliable label information, particularly in scenarios with large Non-Line of Sight diameters, leading to positioning errors greater than 10 meters.

Innovation Solution

A method for acquiring label information involves determining a correction parameter to adjust initial label information using actual device information, such as position coordinates or intermediate parameters, to enhance the reliability of the data set for training AI/ML models, thereby improving positioning accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional positioning methods are used in scenarios with large Non-Line of Sight diameters, then the positioning process can be completed, but the positioning accuracy deteriorates with errors greater than 10 meters

Engineering Contradiction:
Improvepositioning accuracyVSAvoidNon-Line of Sight diameter
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent introduces an AI/ML model as an intermediary between the raw positioning measurements and the final position determination. This model learns from labeled training data to recognize patterns and correct errors introduced by NLOS conditions, thereby improving positioning accuracy without being directly affected by the physical NLOS diameter

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent performs preliminary action by collecting and labeling positioning data in advance under various NLOS conditions. This labeled training data is used to train the AI/ML model before actual positioning operations, enabling the model to learn correction patterns proactively rather than reacting to errors in real-time

Inventive Principle:
Principle #10Preliminary action

Solution Approach 3:

The patent implements feedback mechanisms where the AI/ML model continuously learns from the difference between predicted positions and actual positions. This feedback loop allows the model to refine its predictions and improve accuracy over time, compensating for systematic errors caused by NLOS conditions

Inventive Principle:
Principle #23Feedback

2Measurement precision

If AI/ML models are trained with unreliable label information, then the model training process can proceed, but the positioning accuracy deteriorates

Engineering Contradiction:
Improvepositioning accuracyVSAvoidlabel information reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent performs preliminary action by collecting and labeling positioning data in advance under various NLOS conditions. This labeled training data is used to train the AI/ML model before actual positioning operations, enabling the model to learn correction patterns proactively rather than reacting to errors in real-time

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback mechanisms where the AI/ML model continuously learns from the difference between predicted positions and actual positions. This feedback loop allows the model to refine its predictions and improve accuracy over time, compensating for systematic errors caused by NLOS conditions

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4645780A1Method and apparatus for acquiring label information, and device and storage medium
Publication Date: 2025.11.05 GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
  • EP4645780A1 patent drawingFigure 1
  • EP4645780A1 patent drawingFigure 2A~2B
  • EP4645780A1 patent drawingFigure 3~4

AI summary

Provided in the embodiments of the present disclosure are a method and apparatus for acquiring label information, and an electronic device. The method comprises: a first network device determining correction parameters; and the first network device determining, on the basis of the correction parameters, label information corresponding to each data in a data set, wherein the data set and the label information corresponding to each piece of data in the data set are used for training a first model.