Underwater robots have the potential to enhance greatly human ability to explore and utilize the world's oceans. Many research institutions are studying autonomous underwater vehicles (sc AUVs) in order to increase the capabilities of undersea robots. However, the capabilities of current sc AUVs have been largely limited to survey and sample tasks with minimal interaction with the ocean environment.; The ability of untethered robots to manipulate objects in the environment represents a significant advancement towards increasing humankind's capacity to work under the sea. The work presented in this thesis establishes a new class of underwater robotic systems with intervention capabilities--the Intervention Autonomous Underwater Vehicle (I sc AUV). Through this research, in a joint program between the Stanford University Aerospace Robotics Laboratory (A sc RL) and the Monterey Bay Aquarium Research Institute (M sc BARI), O scTTER--an Ocean Technology Testbed for Engineering Research, the world's first I sc AUV, has been developed to investigate experimentally supporting technologies that will enable this new class of underwater robots.; In this work, the general principles of Object-Based Task-Level Control, where human perception and judgment are intelligently merged with computer computation and control, have been adopted in the design approach. In doing so, the constraints on communication between the human and robot are reduced, and more importantly, the capabilities of the resulting human/robot team are beyond that of either manual teleoperation or total autonomy alone.; The philosophy of concurrently specifying a programming architecture with software tools is introduced as a methodology for creating shareable architectures. A generic programming architecture for underwater robots is a direct contribution of this research. This architecture can be easily adopted by other robotic programmers by using an associated set of off-the-shelf software tools that have been assembled to create components fitting together in an unified framework.; Specifically to support intervention, a practical stereo-vision sensing system was created to locate underwater objects, and an unique approach called Constrained Relative Trajectories was developed to control the robot without occluding the vision system. Integrated together within the O sc TTER I sc AUV, the design approach and the newly-developed technologies contributed by this thesis are validated experimentally in the first-ever demonstration of a multi-phased intervention mission--where O sc TTER successfully searches for, finds and approaches, docks to and retrieves an object sitting at the bottom of a test tank upon high-level command of a human supervisor.
展开▼
