3D-belief space planning for underwater mobile grasping

In the present work, a method based on belief space planning, assuming maximum likelihood of the observations, is applied to the planning of manipulation for an underwater robotic arm. The manipulator is rigidly connected to a floating platform, such as a ROV (Remotely Operated Vehicle) or an AUV (Autonomous Underwater Vehicle). The arm and platform motions are statistically independent from the motion of the object to be grasped. The belief space planning strategy allows to plan the reaching movement with the objectives of concurrently reducing the end-effector distance from the target and the uncertainty on the measure. In this paper the implementation of belief space planning assuming maximum observation likelihood is reported and is applied to an industrial class of underwater arm modelled as purely deterministic. Results obtained by the proposed strategy within preliminary experiments in air are here provided.