AI Positioning Model for Indoor NLOS Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current positioning methods in NR systems, especially in indoor scenarios with severe multipath and non-line-of-sight conditions, struggle to achieve high-precision positioning due to errors caused by large non-line-of-sight probabilities and limited training data for diverse scenarios, leading to inconsistent positioning accuracy.
Innovation Solution
The implementation of an AI-based positioning method using neural networks that processes measurement information from multiple base stations to determine location, allowing for feedback and model updates, enabling high-precision positioning across various scenarios by constructing and transferring training data sets to adapt to changing environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional time measurement-based or angle-based positioning methods are used in NR systems, then the positioning can be implemented with current technology, but the positioning precision is insufficient especially in indoor scenarios with severe multipath and non-line-of-sight conditions
Solution Approach 1:
The patent replaces traditional geometric positioning methods (time-based TDOA or angle-based methods) with an AI-based neural network system. The neural network learns complex patterns from channel measurement information and directly predicts position coordinates, substituting the mechanical/geometric calculation approach with an intelligent learning-based system that can handle multipath and NLOS conditions more effectively.
Solution Approach 2:
The patent changes the input parameters from direct measurement values (time of arrival, angle of arrival) to comprehensive channel measurement information including channel impulse response, reference signal received power, and reference signal received quality. This parameter transformation enables the neural network to capture more environmental characteristics and improve positioning accuracy in challenging conditions.
2Ease of manufacture
If positioning models are trained with limited data for specific scenarios, then the model can be trained and deployed, but the positioning accuracy becomes inconsistent when applied to diverse scenarios with varying environmental conditions
Solution Approach 1:
The patent creates a universal positioning model that can handle multiple scenarios through transfer learning. The pre-trained model serves as a foundation that can be adapted to different environments (indoor, outdoor, urban, rural) by training on scenario-specific data. This multi-functional approach allows a single model architecture to serve diverse positioning needs while maintaining consistency and accuracy across different conditions.
Solution Approach 2:
The patent implements preliminary training of the neural network model on large-scale diverse data before deployment. This pre-training phase enables the model to learn general positioning patterns and features that are transferable across different scenarios. When deployed in specific scenarios, the model can be fine-tuned with limited scenario-specific data, reducing the need for extensive scenario-specific training data while maintaining high accuracy.
Data Source
AI summary
Provided are a positioning method a model training method, and a first device. The positioning method includes: sending, by a first device, measurement information; and receiving by the first device, location information of the first device. The location information is obtained by processing the measurement information based on a positioning model.