Underwater Vehicle Positioning by Correntropy-Based Fuzzy Multi-Sensor Fusion.

Area of Science:

• Robotics
• Marine Engineering
• Signal Processing

Background:

• Accurate underwater vehicle positioning is critical for mission success.
• Kalman filtering is standard for multi-sensor fusion but sensitive to noise covariance and outliers.
• Changing underwater environments and multipath effects introduce significant challenges.

Purpose of the Study:

• To develop an adaptive multi-sensor fusion method for robust underwater vehicle positioning.
• To enhance Kalman filter performance in dynamic and non-Gaussian noise conditions.
• To address the limitations of conventional methods in handling outliers and changing noise covariances.

Main Methods:

• An adaptive multi-sensor fusion technique using information-theoretic, learning-based fuzzy rules for Kalman filter covariance adaptation.
• Novel correntropy-based metrics (Gaussian and Versoria kernels) for theoretical and actual covariance matching.
• Integration of correntropy metrics with fuzzy logic for outlier robustness.

Main Results:

• The proposed method demonstrates robustness against outliers in nonlinear dynamic underwater environments.
• Monte-Carlo simulations show substantial improvements in underwater position estimation accuracy.
• Effective adaptation of process and measurement noise covariance under varying conditions.

Conclusions:

• The developed adaptive fusion method significantly enhances underwater vehicle positioning accuracy and reliability.
• The combination of correntropy and fuzzy logic provides a powerful solution for handling non-Gaussian outliers.
• This approach offers a more dependable solution for underwater navigation in challenging conditions.