.png){kind=icon}
.png){kind=icon}
Return
Multisensor Gaussian-Cauchy Kernel Maximum Correntropy KF With Adaptive Kernel Bandwidth
.png){kind=icon}
PRE
AI
DOI：10.1109/lsp.2026.3709726.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The maximum correntropy criterion (MCC) based Kalman filter (KF) enhances estimation robustness against non-Gaussian noise, but its Gaussian kernel suffers from limited suppression of heavy-tailed noise/outliers and high sensitivity to kernel bandwidth. To address these issues, this letter proposes an improved multisensor Gaussian-Cauchy kernel-based maximum correntropy KF (GCMCKF) with adaptive kernel bandwidth. Unlike the MCC-based KF which requires a fixed-point iteration, the proposed GCMCKF is derived directly into a KF-like form. It utilizes the measurement residuals and state prediction errors to construct a Gaussian-Cauchy kernel cost function, balancing small-error accuracy with large-error robustness. Furthermore, a dual-channel kernel bandwidth update strategy is introduced, which uses a sliding window to adjust the two kernel bandwidths independently. Simulation results for bearing-only target tracking demonstrate the proposed algorithm achieves superior performance in non-Gaussian noise.
Keywords:
Gaussian-Cauchy kernel
maximum correntropy criterion
Kalman filter
adaptive kernel bandwidth
non-Gaussian noise
Journal
I
IF:
3.9
Papers:
610
Citations:
0