Seamless MEMS-INS/Geomagnetic Navigation System Based on Deep-Learning Strong Tracking Square-Root Cubature Kalman...

Area of Science:

• Navigation Systems Engineering
• Signal Processing
• Artificial Intelligence in Navigation

Background:

• Microelectromechanical systems-inertial navigation systems (MEMS-INS) suffer from drift, limiting their navigation accuracy.
• Geomagnetic navigation systems (MNS) can aid INS but are unreliable in geomagnetically unlocked environments.
• Seamless and robust navigation is crucial for applications relying on MEMS-INS/MNS.

Purpose of the Study:

• To propose a hybrid seamless MEMS-INS/MNS strategy to suppress drift and improve navigation continuity.
• To enhance the robustness and computational efficiency of integrated navigation systems.
• To achieve high-precision navigation estimation even during periods of MNS signal loss.

Main Methods:

• A hybrid seamless MEMS-INS/MNS strategy combining a strongly tracked square-root cubature Kalman filter (STSRCKF) with deep self-learning (DSL) was developed.
• The method establishes a relationship between deep Kalman filter gain and optimal estimation, incorporating strong tracking and square-root filtering with singular value decomposition.
• A nonlinear autoregressive neural network (NARX) with exogenous inputs was introduced for deep self-learning capabilities.

Main Results:

• The ST-SRCKF method achieved a heading accuracy error of 1.29°, improving accuracy by 90.10% over single INS and 9.20% over traditional integrated navigation.
• The DSL-STSRCKF method maintained a heading accuracy of 1.33° even during MNS lockout periods, an 89.80% improvement over single INS.
• The proposed strategy demonstrated improved robustness and computational efficiency for continuous high-precision navigation estimation.

Conclusions:

• The DSL-STSRCKF strategy effectively suppresses MEMS-INS drift and enhances seamless navigation capabilities in challenging environments.
• The integration of deep self-learning significantly improves heading accuracy and robustness, particularly during MNS signal outages.
• This hybrid approach offers a promising solution for continuous, high-precision navigation estimation in MEMS-INS/MNS applications.