Neural Network-Based Self-Tuning PID Control for Underwater Vehicles.

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Proportional + Integral + Derivative (PID) controllers are widely used but often require manual tuning.
• Traditional PID tuning methods are typically offline and struggle with systems experiencing continuous parameter changes.
• Underwater Remotely Operated Vehicles (ROVs) present a challenge due to varying parameters (e.g., payload, buoyancy) and environmental disturbances.

Purpose of the Study:

• To develop an online adaptive control strategy for PID controllers in underwater ROVs.
• To address the performance degradation of fixed-gain PID controllers in dynamic underwater environments.
• To propose a Neural Network (NN)-based auto-tuning mechanism for PID gains.

Main Methods:

• Implementation of a Neural Network (NN) to automatically estimate optimal PID controller gains.
• Online adjustment of PID gains by the NN to minimize position tracking errors.
• Simulation of the proposed auto-tune PID controller on an underactuated 6-DOF underwater ROV model.
• Real-time experimental validation on an underactuated mini ROV.

Main Results:

• The NN-based auto-tune PID controller successfully estimated suitable gains online, ensuring system stability.
• The proposed method demonstrated a significant reduction in position tracking errors compared to fixed-gain controllers.
• Simulations and real-time experiments confirmed the effectiveness and robustness of the adaptive control scheme.

Conclusions:

• The Neural Network-based auto-tune PID controller offers an effective solution for adaptive control of underwater ROVs.
• Online gain tuning is crucial for maintaining ROV performance amidst changing operational conditions and environmental factors.
• This approach enhances ROV stability and precision, paving the way for more autonomous underwater operations.