System Modeling and Adaptive Strong-Tracking Filter Design for Underwater Polarization Navigation Against Variable Measurement Bias

Polarization navigation (PN) is an emerging biomimetic navigation method. It has the characteristics of passivity, full autonomy and free of error accumulation, providing a new solution for underwater navigation. However, complex underwater environments and vehicle motion may cause polarization measurement bias. To improve the reliability against variable measurement bias, a robust navigation system integrating polarization and inertial information has been developed. First, a novel polarization observation model with measurement bias in sun vector is established. Modulus constraint about raw specific force is introduced to improve the applicability of gravity observation model during acceleration motion. Meanwhile, an asynchronous fusion strategy has been designed to handle the mismatch of data update rates between different navigation methods. Second, an adaptive two-stage strong-tracking Kalman filter is designed to ensure fast response capability of the system to variable measurement bias. Finally, experimental results demonstrate the superiority of the method over existing observation models and filters in real underwater environments. This article provides an effective solution for improving PN performance in unfamiliar and complex underwater environments.