A Novel Variational Bayesian Method with Unknown Noise for Underwater INS/DVL/USBL Localization.

Area of Science:

• Robotics and Control Systems
• Signal Processing
• Ocean Engineering

Background:

• Accurate state estimation is crucial for underwater systems, but challenging due to unpredictable system noise.
• Traditional methods struggle with uncertain noise models, leading to reduced precision in state determination.
• Underwater environments present unique difficulties in obtaining reliable prior information about system noise.

Purpose of the Study:

• To develop a novel adaptive cubature Kalman filter for improved state estimation in complex underwater environments.
• To address the challenge of uncertain system noise by employing an inverse-Wishart distribution.
• To enhance the characterization of system noise dynamics and uncertainty in underwater applications.

Main Methods:

• Proposed an inverse-Wishart (IW) based variational Bayesian adaptive cubature Kalman filter (IW-VACKF).
• Utilized the inverse-Wishart distribution as a conjugate prior for system noise covariance matrices.
• Introduced a mixing probability vector to model the uncertainty and dynamics of state noise.
• Derived state transition and measurement processes as hierarchical Gaussian models.
• Employed variational Bayesian methods for joint posterior information calculation.

Main Results:

• The IW-VACKF demonstrated improved state estimation precision in simulations.
• Real-world trials confirmed the filter's effectiveness in complex underwater conditions.
• The proposed method efficiently handles uncertain system noise, outperforming conventional approaches.

Conclusions:

• The IW-VACKF offers a robust solution for precise state estimation in challenging underwater scenarios.
• Effective characterization of system noise using inverse-Wishart distribution and mixing probabilities is key to improved accuracy.
• The developed filter provides a significant advancement for underwater navigation and control systems.