Pose Estimation of Unmanned Underwater Vehicles Using Augmented Reality Marker-Based Simulations

Area of Science:

• Robotics
• Computer Vision
• Control Systems

Background:

• Accurate pose estimation is crucial for Unmanned Underwater Vehicle (UUV) navigation.
• Monocular vision systems offer a cost-effective solution but face challenges in underwater environments.
• Integrating perception and control is essential for robust UUV operation.

Purpose of the Study:

• To present a simulation-based framework for UUV pose estimation using monocular vision.
• To evaluate the performance of a Perspective-n-Point (PnP) algorithm and a Proportional-Integral-Derivative (PID) controller in a ROS-Gazebo environment.
• To assess the system's robustness against hydrodynamic disturbances.

Main Methods:

• Utilized the RexRov2 UUV model in ROS-Gazebo with ArUco_ROS for marker detection.
• Employed a Perspective-n-Point (PnP) algorithm for pose estimation.
• Implemented a PID controller for motion regulation based on visual features.

Main Results:

• Achieved stable and accurate pose estimation under nominal conditions, with close agreement to ground truth.
• Demonstrated smooth trajectory tracking in controlled simulated environments.
• Observed significant deviations and instability in tracking performance under increased hydrodynamic disturbances.

Conclusions:

• The proposed framework is a valuable tool for testing underwater navigation and perception-control integration.
• Classical PID control demonstrates limitations in nonlinear underwater environments with disturbances.
• Further research into robust control strategies is necessary for reliable UUV operation in challenging conditions.