Powered by OpenAIRE graph
Found an issue? Give us feedback
addClaim

Machine learning techniques for multi-GNSS positioning

Authors: Lee, Pin-Hsun;

Machine learning techniques for multi-GNSS positioning

Abstract

Les systèmes mondiaux de navigation par satellite (GNSS) sont essentiels pour fournir des informations de position, vitesse et temps (PVT) dans un large éventail d'applications, notamment dans les domaines du transport et des services basés sur la localisation. Toutefois, les environnements urbains posent d'importants défis au positionnement GNSS en raison des interruptions de signal, des réceptions sans visibilité directe (NLOS), ainsi que des effets multitrajets causés par les immeubles de grande hauteur et les rues étroites. Bien que l'utilisation de GNSS multi-constellations augmente efficacement le nombre de signaux disponibles, l'inclusion de signaux dégradés peut fortement détériorer la précision du positionnement. Cette étude propose un algorithme de moindres carrés pondérés (WLS) assisté par apprentissage automatique, associé à un nouveau cadre de fonctions d'activation, afin d'améliorer les performances du positionnement. Les algorithmes d'apprentissage ensemblistes sont entraînés à partir de plusieurs indicateurs de qualité du signal afin d’identifier les signaux de faible qualité et de leur attribuer un score de qualité. Les fonctions d'activation sont ensuite employées pour convertir les scores prédits par l'apprentissage automatique en poids appropriés pour la solution WLS. Différentes fonctions d'activation sont explorées et analysées, parmi lesquelles la fonction sigmoïde offre systématiquement les améliorations les plus significatives pour différents résultats issus de l'apprentissage automatique et différentes configurations de constellations GNSS. Les résultats expérimentaux utilisant des ensembles de données réels démontrent des réductions substantielles des erreurs de positionnement pour les scénarios à constellation unique et multi-constellations. De plus, nous montrons que l'utilisation de plusieurs constellations associée aux algorithmes proposés améliore non seulement la précision du positionnement, mais également la disponibilité temporelle. Ce travail met en évidence les avantages d'intégrer des algorithmes d'apprentissage automatique à un cadre de fonctions d'activation pour obtenir des solutions améliorées de navigation par satellite

Global Navigation Satellite Systems (GNSS) are essential for delivering position, velocity, and time (PVT) information in a wide range of applications, including transportation and location-based services. However, urban environments pose significant challenges to GNSS positioning due to signal outages, non-line-of-sight (NLOS) reception, and multipath effects caused by high-rise buildings and narrow streets. While multi-constellation GNSS effectively increases the number of available signals, the inclusion of degraded signals can severely deteriorate positioning accuracy. This study proposes a machine learning-aided weighted least squares (WLS) algorithm with a novel activation function framework to enhance positioning performance. Ensemble learning algorithms are trained on several signal indicators to identify low-quality signals and provide quality scores. Activation functions are then employed to map the predicted scores from machine learning into appropriate weights for the WLS solution. Various activation functions are explored and analyzed, among which the sigmoid function consistently delivers the largest improvements with different machine learning outputs and GNSS constellation configurations. Experimental results using real-world datasets demonstrate substantial reductions in positioning errors for both single- and multi-constellation scenarios. Furthermore, we showed that using multi-constellations with the purposed algorithms not only improves the positioning accuracy but also time availability. This work shows the advantage of integrating machine learning algorithms with an activation function framework for improved satellite navigation solutions

Keywords

Electrical and Computer Engineering

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Upload OA version
Are you the author of this publication? Upload your Open Access version to Zenodo!
It’s fast and easy, just two clicks!