Roy M. Turner
Department of Computer Science
University of New Hampshire
Durham, NH 03824
(603)862-2980
rmt@unh.edu
We have entered a time when humanity's egregious misuse of environmental resources threatens to destroy or at least severely damage our civilization. Fortunately, this coincides with (or possibly has caused) an increased awareness in society of the need to take action, as well as with the technological sophistication to possibly ameliorate at least the worst of our impact on the environment (and vice versa). To do this, however, we must first understand the environment.
Crucial to understanding the environment is understanding the ocean. The
ocean has a tremendous, perhaps predominant, influence on the global energy
and CObalances and on the global ecosystem. Unfortunately, as Blidberg
##1[][]blidberg91 has stated, we know less about the deep ocean than we
do about the surface of Neptune.
There are several reasons for this. First, the ocean is opaque to most sensors except sonar. This means that orbital data about the deep ocean is nearly impossible to obtain. Sonar is imprecise, noisy, and gives little information apart from distance and shape. Thus the amount and quality of information obtainable from the surface or shallow water about the deep ocean is severely limited. Second, the ocean is an extremely hostile environment for the kind of long-term human presence that would be useful in gathering data needed to understand it. Deep submersibles, such as Alvin, are few and extraordinarily expensive to use, and they are limited in the length of time they may remain on station; deep-ocean habitats are nonexistant. Third, it is difficult to put in place, communicate with, and retrieve instruments in the deep ocean, in many ways more difficult than with space-borne instruments. Though remotely-operated vehicles (ROVs), such as the Woods Hole Oceanographic Institute's JASON and others, can be used to return valuable scientific data from the (relatively) deep ocean, they require the presence of a costly support vessel on-station for extended periods of time.
The best hope for understanding the ocean is autonomous underwater vehicles (AUVs). AUVs can serve as avatars for human scientists, providing an indirect human presence in the ocean for ocean science data gathering, or providing on-site spot-checks to ``ground truth'' a satellite's data. They need little in the way of a support vessel. Between launch and recovery, the support vessel can return to base. In fact, in some situations, no support vessel is needed, as they can be launched from shore. They are also inexpensive with respect to an oceanographic vessel and much more expendable, if necessary.
AUVs are beginning to be used in ocean science research to good advantage. For example, WHOI's ABE (the Autonomous Benthic Explorer) [\protect\citenameYoerger et al., 1991] has been used to study hydrothermal vents for relatively long periods of time, and MIT Sea Grant's Odyssey AUV [\protect\citenameBellingham et al., 1993] has been used for autonomous photography missions under pack ice in Antarctica. The Marine System Engineering Laboratory (MSEL) at Northeastern University is building a long-range AUV (LRAUV) for ocean science missions [\protect\citenameMSEL, 1991]. Such vehicles, like ABE, could remain on-station for extended periods of time, functioning as ``underwater satellites'' [\protect\citenameBlidberg et al., 1991]-returning large quantities of data at a relatively low cost per byte.
For all but the most routine missions, AUVs need intelligent control software. For long-range AUVs, the mission controller must be able to create and modify mission plans, manage resources, communicate with others (to transmit data, to coordinate with other AUVs for large area surveys, to coordinate with satellites for ground truthing data, etc.), and handle unanticipated events that threaten the mission or the vehicle.
The Orca project at the University of New Hampshire focuses on creating a robust, intelligent controller for long-range ocean science AUVs. Orca [\protect\citenameTurner &Stevenson, 1991][\protect\citenameTurner, 1994] is a schema-based, context-sensitive, adaptive reasoner that will control MSEL's long-range and short-range AUVs for a variety of ocean science missions.
In this paper, we first discuss the long-range AUV that will be controlled by Orca and requirements for controlling such a vehicle for ocean science missions. We then describe the Orca program in some detail. We conclude with a discussion of our current status and future plans.