GNSS Receiver Satellite Subset Selection for Multipath Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
GNSS receivers experience significant position and velocity errors due to multipath effects in urban canyons, where signals are reflected off surrounding structures, leading to distorted pseudorange and Doppler measurements, which can result in inaccurate navigation solutions.
Innovation Solution
Implementing machine learning techniques, specifically using a RANSAC algorithm and neural networks, to select a subset of GNSS satellites based on features such as RANSAC residuals and other quality metrics, reducing the impact of multipath reflections by computing position and velocity solutions from these selected satellites.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all visible GNSS satellites are used for position calculation, then the quantity of satellites increases, but position accuracy deteriorates due to multipath effects from reflected signals
Solution Approach 1:
The patent segments the set of all visible GNSS satellites into two subsets: selected satellites (with reliable line-of-sight signals) and excluded satellites (with multipath-contaminated signals). This segmentation allows the system to use only the quality subset for position calculation, resolving the contradiction between quantity and accuracy by selectively including only beneficial satellites.
Solution Approach 2:
The patent applies local quality assessment by evaluating individual satellite signal characteristics (RANSAC residuals, signal-to-noise ratio, elevation angle) to determine which satellites have high quality line-of-sight signals versus low quality multipath signals. Each satellite is assessed locally rather than treating all satellites uniformly, enabling accurate selection of high-quality signals for position calculation.
2Measurement precision
If machine learning techniques are implemented to select satellite subsets, then position accuracy improves, but device complexity increases
Solution Approach 1:
The patent implements preliminary action by pre-training machine learning models (neural networks, random forests, gradient boosting) offline to learn the complex relationship between satellite signal features and position accuracy. The pre-trained models are then deployed in the GNSS receiver, where they perform rapid inference without requiring complex real-time training computations, thus improving accuracy while limiting real-time processing complexity.
Solution Approach 2:
The patent substitutes traditional mechanical/statistical signal processing methods with machine learning-based selection. Instead of using conventional threshold-based or geometric selection algorithms, the system employs trained ML models that automatically learn optimal selection criteria from training data, replacing complex manual tuning with adaptive intelligent decision-making.
Data Source
AI summary
Machine learning techniques are used to perform RANSAC like processing in a GNSS receiver. A model (e.g., one or more neural networks) is trained to perform this processing to generate a selection of a subset of GNSS SVs. In one embodiment, the trained model is used during inferencing in a GNSS receiver. A method in a GNSS receiver can include the following operations: receiving GNSS signals from a plurality of SVs; extracting a set of features from the received GNSS signals, the set of features being predetermined based on a trained model in the GNSS receiver, the trained model having been trained to select a subset of GNSS SVs based on the set of features including RANSAC (random sample consensus) residuals; applying the set of features as an input to the trained model; generating, by the trained model, a selection of a subset of GNSS SVs based in part on RANSAC residuals and based on other features in the set of features; and computing a position solution using GNSS signals received from the subset of GNSS SVs.