Observer-Based Adaptive Neural Network Trajectory Tracking Control for Remotely Operated Vehicle

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Remotely Operated Vehicles (ROVs) often face challenges with unknown dynamic models and unmeasured states, hindering precise trajectory tracking.
• Existing control methods struggle with the complexities of inaccurate thrust models and unmeasured velocity/angular velocity states in real-world ROV systems.

Purpose of the Study:

• To develop an adaptive trajectory tracking control scheme for ROVs that addresses unknown dynamics, unmeasured states, and inaccurate thrust models.
• To enhance the accuracy and robustness of ROV navigation through advanced control strategies.
• To guarantee finite-time convergence of trajectory tracking errors in challenging ROV operational environments.

Main Methods:

• A novel local recurrent neural network (local RNN) with fast learning is proposed for online identification of the unknown ROV dynamic model.
• An adaptive terminal sliding-mode state observer, integrated with the local RNN, estimates unmeasured velocity and angular velocity states.
• An adaptive scale factor is introduced to compensate for thrust model inaccuracies, with the thruster control signal directly used as system input.

Main Results:

• The proposed adaptive control law, utilizing local RNN outputs, adaptive scale factor, and state estimates, ensures robust trajectory tracking.
• Finite-time convergence of the trajectory tracking error is mathematically guaranteed by the adaptive terminal sliding-mode state observer.
• Simulations demonstrate the effectiveness and stability of the developed adaptive trajectory tracking control scheme for ROVs.

Conclusions:

• The novel adaptive control strategy effectively handles unknown dynamics, unmeasured states, and thrust model uncertainties in ROV systems.
• The integration of local RNNs and adaptive observers provides a robust solution for precise ROV trajectory tracking.
• The proposed method offers a significant advancement for ROV control, validated through simulation studies.