Enhanced EKF-based Time Calibration for GNSS/UWB Tight Integration

Tight integration of low-cost Ultra-Wide Band (UWB) ranging sensors with mass-market Global Navigation Satellite System (GNSS) receivers is gaining attention as a high-accuracy positioning strategy for consumer applications dealing with challenging environments. However, due to independent clocks em...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE sensors journal 2023-01, Vol.23 (1), p.1-1
Hauptverfasser: Guo, Yihan, Vouch, Oliviero, Zocca, Simone, Minetto, Alex, Dovis, Fabio
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Tight integration of low-cost Ultra-Wide Band (UWB) ranging sensors with mass-market Global Navigation Satellite System (GNSS) receivers is gaining attention as a high-accuracy positioning strategy for consumer applications dealing with challenging environments. However, due to independent clocks embedded in Commercial-Off-The-Shelf (COTS) chipsets, the time scales associated with sensor measurements are misaligned, leading to inconsistent data fusion. Centralized, recursive filtering architectures can compensate for this offset and achieve accurate state estimation. In line with this, a GNSS/UWB tight integration scheme based on an Extended Kalman Filter (EKF) is developed that performs online time calibration of the sensors' measurements by recursively modeling the GNSS/UWB time-offset as an additional unknown in the system state-space model. Furthermore, a double-update filtering model is proposed that embeds optimizations for the adaptive weighting of UWB measurements. Simulation results show that the double-update EKF algorithm can achieve a horizontal positioning accuracy gain of 41.60% over a plain EKF integration with uncalibrated time-offset and of 15.43% over the EKF with naive time-offset calibration. Moreover, a real-world experimental assessment demonstrates improved Root-Mean-Square Error (RMSE) performance of 57.58% and 31.03%, respectively.
ISSN:1530-437X
1558-1748
DOI:10.1109/JSEN.2022.3223974