A Novel Variational Bayesian Method Based on Student's t Noise for Underwater Localization

Area of Science:

• Robotics and Control Systems
• Signal Processing
• Underwater Acoustics

Background:

• Underwater navigation systems are susceptible to multipath effects from acoustic sensors, causing measurement outliers.
• These outliers degrade the accuracy of integrated navigation state estimation.
• Existing Kalman filter methods struggle with heavy-tailed noise distributions common in underwater acoustics.

Purpose of the Study:

• To propose a robust state estimation method for underwater integrated navigation.
• To improve accuracy and robustness against measurement outliers caused by multipath effects.
• To mitigate the impact of unknown noise characteristics on navigation accuracy.

Main Methods:

• Developed a sliding window variational Kalman filter based on Student's t-distribution (SWVKF-ST).
• Utilized Student's t-distribution to model heavy-tailed noise and inverse Wishart distribution for noise covariance prior.
• Employed a variational Bayesian framework for joint estimation of state variables and measurements within a sliding window.
• Incorporated multiple fading factors to prevent degradation of estimation accuracy.

Main Results:

• The SWVKF-ST demonstrated superior filtering precision and robustness to outliers compared to conventional filters in simulations and experiments.
• The method effectively reduced uncertainty in the measurement noise covariance matrix.
• Validated excellent adaptability of the SWVKF-ST in complex underwater environments.

Conclusions:

• The SWVKF-ST is a highly effective method for improving state estimation accuracy in underwater integrated navigation.
• The proposed filter offers enhanced robustness against outliers and unknown noise characteristics.
• This approach significantly advances the reliability of acoustic sensing for underwater navigation.