5G CIR Positioning Model Training for Higher Location Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of achieving precise positioning in communication systems, particularly with the widespread deployment of 5G network devices, has emerged as an urgent issue.
Innovation Solution
A method and apparatus for training a positioning model using channel impulse response (CIR) data, involving preprocessing and conversion of CIR data into a format suitable for convolutional neural networks (CNN) or K-nearest neighbor (KNN) models, to enhance positioning accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional positioning methods are used in 5G networks, then network deployment is simplified, but positioning accuracy is insufficient
Solution Approach 1:
The patent applies preliminary action by pre-collecting channel impulse response data from multiple reference points and pre-training positioning models offline before actual positioning operations. The CIR data is preprocessed and stored in a database, and multiple positioning models are pre-trained with different parameters. During actual positioning, the system simply selects and applies the appropriate pre-trained model, avoiding complex real-time calculations while achieving high accuracy.
Solution Approach 2:
The patent utilizes parameter changes by training multiple positioning models with different parameters (such as different network structures, learning rates, or architectural configurations) and selecting the optimal model based on performance metrics. This allows the system to adapt to varying positioning scenarios and environmental conditions without increasing operational complexity, as the parameter optimization is completed during the offline training phase.
2Measurement precision
If more reference signals are used for positioning, then positioning accuracy improves, but signal processing complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-processing channel impulse response data from multiple reference signals during the offline phase. The CIR data is collected, processed, and stored in advance, transforming complex signal processing tasks into pre-computed results that can be directly applied during positioning operations without requiring complex real-time signal analysis.
Solution Approach 2:
The patent uses copying by creating multiple positioning models that replicate the positioning function with different parameter configurations. Instead of processing all reference signals through a single complex model in real-time, the system creates simplified copies of positioning models that can be selectively applied, reducing the computational burden during actual positioning while maintaining accuracy through model selection.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The present disclosure provides a method for training a positioning model, a method for positioning, an apparatus for training a positioning model, and an apparatus for positioning. The method for training the positioning model includes: obtaining first channel impulse response (CIR), where the first CIR is obtained based on measurement of a reference signal of a first reference point; merging the first CIR and a first location label to form first label data; and training the positioning model based on the first label data. The reference signal is used for positioning, the reference signal is sent or received by one or more communication devices, and the one or more communication devices include a first communication device. In a 5G communication system, the CIR of the reference signal is easily collected. Thus, a large amount of CIR data can be obtained. A more accurate positioning model may be trained based on the large amount of CIR data. That is, when the positioning model is used for positioning, the accuracy of the obtained positioning result is greatly improved.