GNSS Positioning via Dual Neural Networks for Interference Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Global navigation satellite systems (GNSS) face challenges in achieving precise positioning without fixed ground-based reference stations, particularly in environments with signal interference, which limits their accuracy and reliability.
Innovation Solution
The method employs two trained artificial neural networks to process GNSS raw data, where the first network identifies low-quality data and the second network improves it, combining the outputs to provide a corrected data set for precise object positioning, thereby enhancing GNSS accuracy without the need for fixed ground-based reference stations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fixed ground-based reference stations are used to improve GNSS positioning accuracy, then positioning precision is improved, but device complexity and infrastructure requirements increase
Solution Approach 1:
The patent introduces artificial neural networks as intermediary components that process and correct GNSS raw data. The first neural network identifies low-quality data patterns, while the second neural network generates corrections for these identified issues, effectively mediating between raw satellite signals and accurate position calculations without requiring ground reference stations
Solution Approach 2:
The patent creates a virtual reference system by training neural networks on datasets that include both accurate reference positions and corresponding GNSS measurements. The networks learn to replicate the correction functions that would otherwise require physical reference stations, enabling accurate positioning through software-based models rather than hardware infrastructure
2Measurement precision
If two artificial neural networks are used to process GNSS data, then positioning accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent divides the data processing task into two distinct neural network components: the first network specializes in identifying low-quality data characteristics, while the second network focuses on generating corrections based on those identifications. This segmentation allows each network to be optimized for its specific function and reduces the overall computational burden compared to a single comprehensive network
Solution Approach 2:
The first neural network performs a partial analysis by identifying only the low-quality data portions of the GNSS signal, rather than processing the entire dataset comprehensively. This selective approach reduces computational requirements while still enabling the second network to apply targeted corrections where needed
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A method to determine a position of an object (10) based on a GNSS raw data set (RDS), comprising at least the steps of: (S600) feeding the GNSS raw data set (RDS) to a first trained artificial neural network (4) in order to get as output of the first artificial neural network (4) a first data set (FDS) of low quality data and a second data set (SDS) of high quality data, (S700) feeding the first data set (FDS) of low quality data to a second trained artificial neural network (6) in order to get as output of the second artificial neural network (6) an improved data set (IDS), (S800) combining the second data set (SDS) of high quality data and the improved data set (IDS) in order to provide a corrected data set (CDS) indicative for the position of the object (10).