Stanford ARL, in a joint program with the Monterey Bay Aquarium Research Institute (MBARI), is developing technologies that will enable an unmanned, untethered submersible to be a common tool used daily by science and industry in their exploration of the world's ocean. This program is a prime example of the benefits of leveraging the strengths of ARL with the experience of MBARI's engineers and scientists to produce a truly capable vehicle. The development of the OTTER (Ocean Technology Testbed for Engineering Research) vehicle draws upon many the areas which have already been studied by ARL.
In line with the lab philosophy that a human/machine team will perform better than either alone, the OTTER vehicle will be controlled by a person at a high level. The onboard control system is intelligent enough to maintain all system functions and to provide an intuitive, easy-to-use interface to the operator. In this manner, instead of requiring a high skill level brought about by years of training and experience, the end-user, whether a marine scientist or commercial geologist, can direct the vehicle's actions with simple, intuitive commands. This concept of Object-Based Task-Level Control (OBTLC), pioneered by ARL, allows the user to focus his or her attention on decisions based on the manipulation of objects of interest (e.g. rock specimen or sampling equipment) and not on the control the vehicle to manipulate the objects.
In OBTLC, humans usually interface at the top level, doing what they do best-interpreting sensor data, making decisions, and reacting to events-while machines are used to process data and keep tight control of the low-level servos. However, because the system is designed to be vertically integrated, a person will be able to operate the system at all levels, perhaps directly controlling a thruster to help maneuver the vehicle out of a crevice.
This program is truly a microcosm of the lab. Research is being accomplished at all levels of control, including the architecture that binds the low servo level to the high task level. At the low level, the precise, high-speed control of a robotic arm underwater is being formulated for use in the autonomous retrieval of undersea specimens. Understanding and then controlling arm/vehicle dynamical interactions are important steps for that task as well. While much research has been performed in the general community on steady-state operations of thruster motors, the understanding of the control of thrusters operating under dynamic conditions is lacking. And it is this understanding which is required to provide thrust for precise positioning of a submersible.
These two basic research areas represent the lab's commitment to develop a basic understanding of the dynamics of electro-mechnical systems and the formulation of controllers to regulate them.
Another area of research in this program focuses on the architecture required to program the vehicle to perform complex tasks autonomously. This research builds upon and will expand the existing structure used throughout the lab to control robotic systems at a task level. Programming a system to react to events and to execute actions associated with events can become quite complex without a formal methodology. Previous and current research take formalisms found in computer science theory and extends them to controlling a real system.
Thus, the OTTER vehicle is not only a testbed for new ideas in systems integration and control, but is also a representative of how the ARL approaches system design with research at all levels of a vertical hierarchy.
- Precise, high-speed underwater manipulators.
- Task-level programming with an advanced finite-state machine paradigm.
- Underwater video navigation, stationkeeping, tracking.
- Coordinated underwater vehicle/manipulator control.
- Optimal control of combined vehicle/camera systems through bandwidth separation.
At Stanford ARL, we have a test tank installation suitable for testing underwater manipulators and thrusters. The test tank is 2.1 m in diameter and 1.5 m deep. A single link manipulator has been outfitted with strain gauges and an accelerometer to both characterize the dynamics of underwater motion and control the arm for high performance slews. An elbow joint has been constructed and will be added to form a two-link underwater manipulator system. For thruster testing, we also have a thrust stand for direct force measurement of thruster performance.
The OTTER vehicle was designed for depths up to 1000 m. It has 6 oil-filled, 1/4 hp brushless direct-current motors for maneuvering and 2 oil-filled, 1 hp brushless drive motors. All motors have an integral electronics stack with an HC11 performing closed-loop torque and velocity control. We have provisions for closed-loop force control using duct-mounted strain gages for future research efforts. On-board batteries provide up to 1.5 kWh of power. While a tether is currently used to trickle charge the batteries, and provide a data link over a serial line, future plans call for tetherless operations using acoustic modems for communications. The sensor suite includes a flux-gate compass, roll-pitch sensor, pressure depth sensor, and an integrated angular-rate/accelerometer package from Systron Donner. The control electronics are centered around two on-board VME cages: one for the vehicle controller, the other for vision processing.
An advanced vision system provides real-time video processing at frame rates from a pair of stereo cameras. This vision system uses sign-correlation hardware and optical flow theory to provide velocity and range maps of the incoming video. Research is being performed to used the correlation of live video with stored video to create a video mosaic of a large region of the ocean floor and then to navigate the vehicle from the stored mosaic.
In addition to our vehicle testbed, we have access to Ventana, a commercially-available tethered ROV, which is used daily by MBARI scientists in the Monterey bay. The eventual transfer of the technologies developed at ARL to MBARI research operations is greatly facilitated by incorporating the experience of MBARI submersible pilots, scientists and engineers early in our experiments.
Demonstrations to Date
Although this program only began in 1991, we have made several important contributions. At the beginning of the program, Tim McLain designed, built and tested a tether tension measuring device that has helped MBARI engineers better understand the tether forces which act on their remotely-operated vehicle (ROV), the Ventana.
We have used the first experimental submersible built by this program, SPOTTER, to show that better overall system performance can be obtained with intelligent compartmentalizing of control based on bandwidth limitations. In this experiment, cameras were mounted on a high-speed pan-and-tilt and used a simple vision processing scheme to track a laser dot. The pan-and-tilt mechanism was itself mounted onto an underwater vehicle. By combining the control of the pan-and-tilt mechanism with that of the vehicle, we were able show that overall tracking performance of the system was superior to either subsystem alone.
A second experimental submersible, OTTER, was built the following year. With this vehicle, we were able to demonstrate the tracking of arbitrary objects underwater. It was during this effort, that we first used optical-flow based, sign-correlation technology to obtain range and velocity maps from a pair of stereo cameras.
Experimental Demonstrations in Progress
A basic control architecture for the task-level control of underwater vehicles has been formulated and research into the particulars of the architecture is ongoing.
In addition, fundamental research is ongoing in the area of hydrodynamics to better understand the interaction between underwater manipulators and the water itself. The culmination of this research will be much improved coordinated control of the vehicle/manipulator system.
Last modified Mon, 1 Nov, 2010 at 14:23