Research on autonomous grasping of an UVMS with model-known object based on monocular visual system

This paper has proposed a method of grasping a model-known object autonomously for an Underwater Vehicle-Manipulator System (UVMS) using monocular vision. The distance between the vehicle and the object is estimated by extracting the pixel radius of the ball-like object, and the vehicle's depth and orientation angle are used to keep the vehicle's station while the manipulator is performing the grasping actions. Other key techniques, including station keeping and grasping strategy, thrust allocation and vehicle control are also presented. Pool experiments demonstrate the validity and practicability of the method.