Many scientific and industrial works that take place
underwater are increasingly demanding for technology to allow
their autonomous execution. Autonomous grasping of unknown
objects by a robotic arm is a task that is being highly studied by
researchers in the last years. This is a really difficult task in itself,
but the problem is bigger if the grasp is done in underwater
environments, where the sensors that can be installed in the arm,
decrease their performance drastically. This paper describes an
approach for semi-autonomous grasping of underwater unknown
objects using a robotic arm. To achieve this purpose, a scan of the
scene is performed using a structured laser beam attached to the
forearm of the manipulator. At the same time, a digital video
camera is used to capture the scene. The laser stripes are
triangulated to obtain a 3D point cloud. The entire process is
shown inside an underwater simulator, acting in this case as a
virtual representation of the real environment. This virtual
representation allows the user to specify the grasp, to see how the
grasp will be executed, and to monitor the final process in real
time. To validate the good performance of the proposed
approach, several grasping experiments with two different objects
have been performed in water tank conditions with a real I-AUV
system.