Predefined-time data-driven distributed neural formation control for underactuated AUV systems

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence
• Marine Engineering

Background:

• Existing data-driven control methods for autonomous underwater vehicles (AUVs) lack interpretability and precise convergence time guarantees.
• Model-free approaches often function as black boxes, limiting understanding and predictability in complex dynamic systems.

Purpose of the Study:

• To develop a novel predefined-time-stable, data-driven modeling framework for AUVs with unknown dynamics.
• To design a distributed neural sliding mode formation tracking controller with a prescribable convergence time.
• To enhance the interpretability and predictability of data-driven control for AUVs.

Main Methods:

• A predefined-time-stable data-driven modeling framework reconstructs nonlinear dynamics online using input-output data.
• Lyapunov analysis is employed to rigorously demonstrate prescribable convergence time bounds.
• Bio-inspired neural shunting dynamics are integrated to mitigate high-frequency chattering in the sliding-mode controller.

Main Results:

• The proposed framework enables online reconstruction of unknown AUV dynamics, reducing reliance on exact model parameters.
• The closed-loop system achieves practical predefined-time stability, with convergence time bounds rigorously proven.
• Simulations confirm the effectiveness and superiority of the proposed data-driven distributed neural sliding mode controller.

Conclusions:

• The developed controller offers a model-free, interpretable, and predictable solution for AUV formation tracking.
• The predefined-time stability framework enhances the reliability and performance of autonomous underwater systems.
• This approach advances data-driven control strategies for complex robotic applications.