Abstract
Optics-based systems may provide high spatial and temporal resolution for close range object detection in underwater environments. By using a monocular camera on a low cost underwater vehicle manipulator system, objects can be tracked by the vehicle and handled by the manipulator. In this paper, a monocular camera is used to detect an object of interest through object detection. Spatial features of the object are extracted, and a dynamic positioning system is designed for the underwater vehicle in order for it to maintain a desired position relative to the object. A manipulator mounted under the vehicle is used to retrieve the object through a developed kinematic control system. Experimental tests verify the proposed methodology. A stability analysis proves asymptotic stability properties for the chosen sliding mode controller and exponential stability for the task error.
Highlights
The need for subsea inspection, maintenance, and repair (IMR) operations in the ocean industries is high, and is expected to increase further in the coming years
This paper studies and develops grasping of a known object using a monocular camera through machine learning and a small underwater vehicle manipulator system (UVMS)
Images containing an object of interest were labeled through a previously developed automatic labeling procedure
Summary
The need for subsea inspection, maintenance, and repair (IMR) operations in the ocean industries is high, and is expected to increase further in the coming years. One of the most recent works in autonomous solutions for UVMS intervention operations is the DexROV project [28], which focuses on reducing the gap between autonomy and tele-operation when controlling ROVs in underwater manipulation operations This project has utilized several technologies for fine manipulation of objects in the water column, such as stereo camera solutions, inertial navigation system, set based task priority control, obstacle avoidance, and a high-end gripper with force/contact sensors [29,30]. (1) Design of a navigation, guidance, and control system for the vehicle to maintain a desired position relative to an object detected through monocular vision and machine learning (2) A stability proof that ensures exponential convergence of the task errors and asymptotic convergence to the sliding mode controller’s sliding surface (3) Experimental testing that proves the effectiveness of the proposed solution for grasping the object with a low cost underwater vehicle manipulator system.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
