Dynamic Factor Precision in Gaussian Belief Propagation: A Tracking Use Case
IEEE Signal Processing Letters · 2026 · 33 · 2061-2065
DOI: 10.1109/lsp.2026.3693204 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Emmanouil Maroulis und 3 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben.
Abstract (Originalsprache)
In this letter, we investigate a consensus-based distributed Kalman filter for single-target tracking over loopy communication graphs using Gaussian belief propagation. Each node (tracker) maintains a local Kalman filter and exchanges Gaussian messages with its neighbors on a factor–graph, implementing a distributed fusion of measurement increments. Instead of the usual static factor precision $L_{ij} = \beta I$, we introduce state-dependent factor precisions that adapt to the local uncertainty and network topology. We further derive simple local bounding rules that guarantee walk-summability of an associated Gaussian surrogate model. In simulations with range-dependent measurement noise and increasingly dense communication graph topologies, a proposed dynamic precision achieves better normalized estimation error square calibration and mean-squared error than static baselines, and remains resilient even in over-connected networks.
Zitieren
APA
Maroulis, E., Hasselmann, K., Flécher, E. L., & Cubber, G. D. (2026). Dynamic Factor Precision in Gaussian Belief Propagation: A Tracking Use Case. IEEE Signal Processing Letters, 33, 2061-2065. https://doi.org/10.1109/lsp.2026.3693204
Vancouver
Maroulis E, Hasselmann K, Flécher EL, Cubber GD. Dynamic Factor Precision in Gaussian Belief Propagation: A Tracking Use Case. IEEE Signal Processing Letters. 2026;33:2061-2065. doi:10.1109/lsp.2026.3693204
BibTeX
@article{Maroulis2026,
author = {Maroulis, Emmanouil and Hasselmann, Ken and Flécher, Emile Le and Cubber, Geert De},
title = {Dynamic Factor Precision in Gaussian Belief Propagation: A Tracking Use Case},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {2061--2065},
doi = {10.1109/lsp.2026.3693204}
}
EUVIMED