Adaptive formation learning control for cooperative AUVs under complete uncertainty.

Area of Science:

• Robotics
• Control Systems Engineering
• Ocean Engineering

Background:

• Adaptive formation control is crucial for Autonomous Underwater Vehicles (AUVs) but often relies on partial knowledge of system dynamics or environmental data.
• Existing methods' reliance on assumptions like known mass matrices limits their adaptability in diverse underwater environments.

Purpose of the Study:

• To develop a configuration-agnostic adaptive formation control framework for AUVs that operates without prior knowledge of system dynamics or environmental conditions.
• To enhance AUV adaptability and robustness in complex underwater scenarios.

Main Methods:

• A two-layer framework is proposed: a cooperative estimator for inter-agent communication and a decentralized deterministic learning (DDL) controller for trajectory control.
• Radial basis function neural networks (RBFNN) are utilized within the framework to store dynamic information, preventing the need for relearning after system restarts.
• The system treats all dynamics, including the mass matrix, as entirely unknown, ensuring broad applicability.

Main Results:

• The framework effectively addresses uncertainties from unknown parameters, unmodeled interactions, and external disturbances like varying currents and pressures.
• Internal mechanisms handle parametric uncertainties and unmodeled interactions, while the controller manages external environmental factors.
• The approach demonstrates enhanced adaptability across diverse underwater environments.

Conclusions:

• Mathematical proofs confirm the stability of the proposed controller.
• Simulation results validate the framework's ability to achieve precise control accuracy and signal boundedness for each agent.
• The study confirms the framework's stability and resilience in complex scenarios, offering a robust solution for AUV formation control.