A Rapid and Adaptive Alignment under Mooring Condition Using Adaptive EKF and CNN-Based Learning.

Area of Science:

• Marine engineering
• Navigation systems
• Signal processing

Background:

• Inertial Navigation System (INS) alignment is crucial for accurate positioning.
• Traditional Extended Kalman Filter (EKF) alignment methods struggle in dynamic mooring environments due to non-stationary zero-velocity measurements.
• Fixed measurement error covariance matrices in EKFs can lead to prolonged alignment times and reduced performance.

Purpose of the Study:

• To develop an improved INS alignment method for dynamic mooring conditions.
• To enhance the accuracy and reduce the alignment time of INS in marine environments.
• To address the limitations of fixed measurement error covariance matrices in EKF-based alignment.

Main Methods:

• Proposed an adaptive Kalman filter integrated with Convolutional Neural Network (CNN)-based learning.
• Adjusted the measurement error covariance matrix dynamically based on mooring conditions.
• Utilized Monte Carlo simulations to evaluate the proposed alignment method.

Main Results:

• The adaptive Kalman filter and CNN-based method demonstrated superior alignment time compared to traditional EKF methods.
• Achieved higher alignment accuracy in simulated mooring environments.
• The dynamic adjustment of the measurement error covariance matrix effectively handled non-stationary conditions.

Conclusions:

• The proposed adaptive INS alignment method offers significant improvements in both speed and accuracy for mooring applications.
• CNN-based learning provides an effective approach to adaptively manage measurement uncertainties in dynamic environments.
• This method enhances the reliability and performance of INS in challenging marine navigation scenarios.