Domain knowledge embedded anti-disturbance autonomous navigation for marine vehicles.

Area of Science:

• Ocean Engineering
• Marine Robotics
• Control Systems

Background:

• Autonomous marine vehicles face navigation challenges due to complex ocean disturbances like wind-wave-current coupling.
• Existing control methods struggle with the unpredictable and composite nature of these environmental factors, impacting safety and precision.

Purpose of the Study:

• To develop an adaptive control framework for enhanced autonomous navigation of marine vehicles.
• To improve the robustness and accuracy of path-following in the presence of complex ocean disturbances.

Main Methods:

• Introduced a domain knowledge-embedded disturbance observation-control framework.
• Fused real-time observation and compensation for composite environmental disturbances using model-free control.
• Embedded domain knowledge into a specialized Kolmogorov-Arnold network to train a machine learning controller.

Main Results:

• Achieved superior adaptability and robustness compared to conventional model-based controllers.
• Demonstrated more accurate path-following and safer operations under complex ocean disturbances.
• Validated effectiveness for both surface vessels and underwater vehicles in offshore wind farm inspection scenarios.

Conclusions:

• The developed framework significantly enhances autonomous navigation capabilities for marine vehicles in challenging ocean environments.
• This approach extends adaptive control theory for marine cyber-physical systems and has broad applications in oceanographic operations.