Abstract
Radar-based tracking of small unmanned aerial vehicles (UAVs) is significantly challenged in urban low-altitude environments by intermittent and multimodal mismatches induced by abrupt target maneuvers and complex propagation effects. Existing GLMB-based trackers, including variational adaptive extensions, primarily address model mismatches through continuous covariance adaptation mechanisms, which become inadequate when heterogeneous and intermittent mismatch forms arise. This paper proposes a mismatch-indicator identification-enabled GLMB (MI-GLMB) filter that extends covariance adaptation by explicitly identifying the underlying mismatch mode and selectively updating only the corresponding mismatch-related covariances. Specifically, a discrete latent variable is introduced to characterize multiple mismatch modes, while continuous latent variables model the associated mode-dependent covariance parameters. A unified variational Bayesian inference framework is formulated to jointly infer the active mismatch mode and estimate its corresponding covariances within the GLMB recursion. Monte Carlo simulations and real-world UAV tracking experiments demonstrate the robustness and accuracy advantages of the proposed MI-GLMB filter in complex urban low-altitude environments.
| Original language | English |
|---|---|
| Article number | 111958 |
| Journal | Aerospace Science and Technology |
| Volume | 175 |
| DOIs | |
| State | Published - Aug 2026 |
Keywords
- Abrupt maneuver
- Bayesian inference
- Generalized labeled multi-Bernoulli filter
- Intermittent multipath deviations,
- Mode and parameter estimation
Fingerprint
Dive into the research topics of 'Mismatch-indicator identification-enabled generalized labeled multi-Bernoulli filter for tracking multi-UAVs in urban environments'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver