Skip to content

Latest commit

 

History

History
104 lines (74 loc) · 3.72 KB

README.md

File metadata and controls

104 lines (74 loc) · 3.72 KB

Reinforcement Learning and Neural-Fuzzy-based Self-Tuned PID Controller for Autonomous Underwater Vehicle (AUV) - MATLAB

This repository showcases a hybrid control system combining Reinforcement Learning (Q-Learning) and Neural-Fuzzy Systems to dynamically tune a PID controller for an Autonomous Underwater Vehicle (AUV). The implementation aims to enhance precision, adaptability, and robustness in underwater environments.

Overview

Autonomous Underwater Vehicles (AUVs) face complex challenges due to the dynamic and uncertain nature of underwater environments. This project integrates:

  • Q-Learning: For adaptive, model-free reinforcement learning.
  • Neural-Fuzzy Logic: For addressing non-linear dynamics and uncertainties in real-time.

The system effectively tunes the PID controller parameters dynamically to improve performance, ensuring stability and energy efficiency in various underwater operations.

Features

  • Dynamic PID Tuning: Ensures optimal performance in real-time.
  • Hybrid Control Architecture: Combines classical PID with modern AI-based approaches.
  • MATLAB Implementation: Detailed simulations and visualization tools.
  • Adaptability: Applicable to other robotic systems like drones and industrial manipulators.

System Architecture

  1. AUV Dynamics Simulation:

    • A 6-DOF model including forces like drag, buoyancy, and thrust.

  2. Reinforcement Learning (Q-Learning):

    • A state-action-reward mechanism for optimal control action selection.

  3. Neural-Fuzzy System:

    • Fuzzy rules combined with neural networks for adaptive control adjustments.

  4. PID Controller:

    • A classical controller enhanced by dynamic parameter tuning.

Repository Contents

  • AUV_Dynamics: MATLAB script for simulating AUV behavior.
  • Q_Learning_Controller: Implementation of Q-Learning for self-tuning the PID controller.
  • Neural_Fuzzy_Tuner: Adaptive control using Neural-Fuzzy logic.
  • `Results_and_Visualization: Tools for analyzing and visualizing system performance.
  • README.md: Documentation for the project.

Prerequisites

  • MATLAB R2021b or later.
  • Required Toolboxes:
    • Control System Toolbox.
    • Fuzzy Logic Toolbox.

Getting Started

Clone the Repository

git clone https://github.com/AlinaBaber/ReinforcementLearning-QLearning-based-self-tuned-PID-controller-for-AUV-MatLab.git
cd ReinforcementLearning-QLearning-based-self-tuned-PID-controller-for-AUV-MatLab

Add to MATLAB Path

Customize Parameters

Modify values in the script files to adjust simulation settings.

Results

The system demonstrates: c5 Comparison-2 c12

  • Improved navigation accuracy.
  • Enhanced stability under disturbances.
  • Reduced energy consumption.
  • Example Visualization

Applications

  • Underwater Robotics: Autonomous navigation for AUVs.
  • Drones: Control and stabilization for UAVs.
  • Industrial Robotics: Adaptive control for manipulators.
  • Future Enhancements
  • Real-world hardware integration for AUV testing.
  • Extending to multi-agent systems.
  • Adapting the framework for surface and aerial robotics.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Contact

For queries or collaboration, reach out to:

Alina Baber

  • 📧 Email: alinababer46@gmail.com
  • 🌐 LinkedIn: Alina Baber Let’s innovate underwater control systems together! 🌊🚀

vbnet This README.md file is structured to be informative and user-friendly while meeting GitHub