AI/ML Positioning Data Augmentation for 5G Mobile Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In 5G New Radio mobile communications, AI/ML positioning models face challenges due to the scarcity and imbalance of training datasets, leading to positioning errors, which data augmentation techniques aim to address by enhancing performance with limited data samples.
Innovation Solution
The proposed solution involves data augmentation methods such as jittering and timing shift transformations, which add noise and timing offsets to original input samples, and Conditional Variational Autoencoding (CVAE) to generate new datasets, expanding the dataset size and improving positioning accuracy without requiring additional data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data collection is performed to obtain training datasets for AI/ML positioning models, then positioning accuracy can be improved, but the overhead and complexity of data collection increases
Solution Approach 1:
The patent applies data augmentation techniques that create synthetic copies of existing training data through transformations such as adding noise, adjusting timing, and generating virtual signal samples. This copying approach allows the system to expand the training dataset without performing additional physical data collection, thereby improving positioning accuracy while avoiding the overhead and complexity associated with extensive field data collection operations
2Measurement precision
If the training dataset size is increased to reduce positioning error, then positioning accuracy improves, but the difficulty of obtaining balanced and complete datasets increases
Solution Approach 1:
The patent employs parameter transformation techniques where existing training data samples are modified by changing parameters such as adding controlled noise, applying timing shifts, and transforming signal characteristics. These parameter changes generate diverse synthetic training samples that expand the dataset size and improve balance without requiring additional difficult-to-obtain field measurements, thus reducing positioning error while avoiding dataset acquisition challenges
3Reliability
If data augmentation techniques are applied to expand training dataset, then positioning model performance improves, but the computational complexity of data processing increases
Solution Approach 1:
The patent implements data augmentation procedures during the offline training phase rather than during real-time positioning operations. By pre-processing and augmenting the training dataset before model training, the system improves positioning model performance while isolating the computational complexity to an offline stage, avoiding real-time processing delays and reducing the apparent complexity during actual positioning operations
Data Source
AI summary
Various solutions for improving positioning by data augmentation for artificial intelligence/machine learning (AI/ML) positioning with respect to an apparatus in mobile communications are described. The apparatus may obtain a data input sample. The apparatus may perform a data augmentation to generate an augmented training data based on the data input sample. The apparatus may perform a model training on a positioning model based on the augmented training data. The apparatus may determine a position information of the apparatus by the positioning model.