OTTER: A Testbed Submersible for Robotics Research

Abstract

Untethered, unmanned submersibles (UUSs) have the potential of becoming useful tools in nuclear reactor inspection and repair and waste-site cleanup and survey, in addition to subsea exploration. Classical "hand-in-glove" control methods used currently by tethered, remotely-operated vehicles (ROVs) are unsuitable for UUS systems which have, at best, limited communications. The Stanford Aerospace Robotics Laboratory (ARL) has joined with the Monterey Bay Aquarium Research Institute (MBARI) in a program to develop new methods of controlling UUS systems. As a part of this effort, a submersible testbed has been built to test and develop new technologies that will make UUSs useful tools for underwater operations. The OTTER, and Ocean Technologies Testbed for Engineering Research, robot was designed to be highly controllable with 8 thrusters for maneuvering and on-board computers for automatic closed-loop control. The sensor suite includes a real-time vision system capable of providing processed information at 30 Hz. Using the OTTER, graduate students have pursued research at all levels of the robotic system. These areas include control achitectures, human-machine interfaces, real-time vision sensing systems, underwater manipulator dynamics and control, and thruster dynamics. This systems approach to studying enabling technologies is one that is fundamental to the research foundation of Stanford ARL's robotic control methodology called "Task-Level Control" (TLC). Within TLC, the operator is recognized as an important member of the human-machine team. The TLC paradigm intelligently blends the decision and judgment capabilities of the human mind with the speed and accuracy of machine computation and control. The following results have been demonstrated experimentally on the OTTER robot: tracking of an arbitrary underwater object, robot station keeping using only vision, automatic creation of mosaics from multiple video images in real time, coordinated arm/vehicle control, and robotic control through a 3D, computer-generated, graphical environment. All of these tasks have potential applications to the inspection an repair of nuclear facilities.