Accurate Time-based Positioning Using Deep Learning With Anchor Selection
Kirmaz A, Sahin T, Michalopoulos DS, Gerstacker W (2025)
Publication Type: Journal article
Publication year: 2025
Journal
Book Volume: 13
Pages Range: 199493 - 199507
DOI: 10.1109/ACCESS.2025.3631399
Abstract
Accurate location information is essential for various emerging applications spanning from industrial to entertainment sectors. However, high-accuracy radio frequency (RF) positioning is hampered by challenging propagation conditions. Artificial neural networks (ANNs) have emerged as enablers for accurate time-of-arrival (ToA) and time-difference-of-arrival (TDoA) estimation by leveraging channel impulse responses (CIRs) to realize high-accuracy ranging. Despite their potential, a comparative evaluation of the positioning performance achieved by ANN-based ToA and TDoA estimation has not been reported yet in the literature. In this paper, we investigate the ranging and positioning performance of ANN-based ToA and TDoA estimation, under constraints on the training dataset size as well as the number of positioning anchors. Additionally, we propose a novel unsupervised method to evaluate the reliability of positioning measurements for the purpose of anchor selection. Based on real-world measurements, we show that ANN-based TDoA estimation yields up to ∼20% higher ranging accuracy than ANN-based ToA estimation when a limited number of channel measurements are available to train the ANNs. The accuracy gap narrows as the amount of training data increases, ultimately resulting in similar accuracy for both schemes in terms of ranging and positioning. Such result serves as a practical guideline for effectively selecting a suitable ANN-based method for positioning based on the amount of available training data. Furthermore, the proposed anchor selection strategy reduces the number of positioning anchors by 25% while either improving the positioning accuracy by up to 10% or maintaining it, depending on the underlying estimator. This reduction is critical in practical deployments as it facilitates energy savings and decreases communication overhead without compromising accuracy.
Authors with CRIS profile
Involved external institutions
Germany (DE)How to cite
APA:
Kirmaz, A., Sahin, T., Michalopoulos, D.S., & Gerstacker, W. (2025). Accurate Time-based Positioning Using Deep Learning With Anchor Selection. IEEE Access, 13, 199493 - 199507. https://doi.org/10.1109/ACCESS.2025.3631399
MLA:
Kirmaz, Anil, et al. "Accurate Time-based Positioning Using Deep Learning With Anchor Selection." IEEE Access 13 (2025): 199493 - 199507.
BibTeX: Download