Autonomous manipulation refers to the capability of a robot system that performs intervention tasks requiring physical contacts with unstructured environments without continuous human supervision. Today, few Autonomous Underwater Vehicles (AUVs) are equipped with autonomous manipulators. SAUVIM (Semi Autonomous Underwater Vehicle for Intervention Mission, University of Hawaii) is one of the first AUVs capable of autonomous manipulation. This paper presents one of the first successful underwater intervention mission of SAUVIM, consisting in a recovery operation of a submerged target. The focus of this manuscript is mainly on the manipulation aspect, introducing the solution adopted in order to achieve the goal.
Most of today's underwater intervention tasks are performed by manned submersibles or remotely operated vehicles (ROVs). In autonomous underwater vehicles, the low bandwidth and significant time delay inherent in acoustic underwater communications are considerable obstacles in remotely operating a manipulation system and reacting to unexpected situations in a timely manner. However, autonomous vehicles with no physical link to the human operator could permit operating in the areas where humans cannot go, such as under ice, in militarily denied areas, or in missions to retrieve hazardous objects. The key element in improving underwater intervention capability in AUVs is autonomous manipulation. It is a challenging technology milestone, which refers to the capability of a robot system that performs intervention tasks requiring physical contacts with unstructured environments without continuous human supervision. Today, few AUVs are equipped with autonomous manipulators. SAUVIM (Semi Autonomous Underwater Vehicle for Intervention Mission, University of Hawaii, Fig. 1) is one of the first AUVs capable of autonomous intervention. Autonomous manipulation systems, unlike teleoperated manipulation systems that are controlled by human operators with the aid of visual and other sensory feedback, must be capable of assessing a situation, including self-calibration based on sensory information, and executing or revising a course of manipulating action without continuous human intervention.
