Stanford Research Communication Program
  Home   Researchers Professionals  About
Archive by Major Area


Social Science

Natural Science

Archive by Year

Fall 1999 - Spring 2000

Fall 2000 - Summer 2001

Fall 2001 - Spring 2002

Fall 2002 - Summer 2003




Underwater Robots: Tools for Scientific Exploration

Andreas Huster
Electrical Engineering
Stanford University
May 2001

Because of the extreme difficulty in gathering subsea observations, very little is known about the oceans and their inhabitants. We do know, however, that the oceans are an extremely complex and equally important part of the world we live in. They cover the majority of the planet's surface, influence our climate, host countless species of plants and animals, are home to important geological processes, and are probably where life itself originated. Because they are notoriously difficult to study, they present one of the final frontiers for exploration.

Because the underwater environment is so dark, much of its biology and geology must be studied at very close range. At greater distances, even powerful lights fail to illuminate the scene sufficiently. The behavior of a jellyfish, for example, can only be observed from a distance of about one meter. These close-up underwater observations, however, would require human divers to descend to dangerous depths, far deeper than is safely possible. Thus, subsea researchers have relied on underwater robots to observe plants and animals, to measure geological processes, to place sensors, and to retrieve samples. Although these robots have demonstrated many impressive capabilities, scientists have been very limited in their ability to explore the oceans with current underwater robot technology.

The goal of this research is to improve underwater robot technology in order to enable more (and more useful) scientific exploration of the oceans. To achieve this goal, underwater robots need, among other new capabilities, more robust and accurate techniques to determine their relative position.

Many tasks in the complex and unknown underwater environment are extremely difficult to achieve with a robot. Human operators, either on-board the robot or connected to it via a tether from a surface vessel, are required to direct its every action and motion. The requirement for a highly trained crew to manage each underwater robot typically drives up the cost of an exploration mission to thousands of dollars a day. This extreme cost of deep ocean research, all just for the equivalent of one pair of eyes or hands, prohibits widespread deployment of robots in the oceans and depresses the current rate of scientific progress.

Furthermore, using human operators to direct an underwater robot is not always feasible. For example, NASA is planning a mission to Europa, a moon of the planet Jupiter, to explore bodies of water believed to exist underneath a layer of surface ice. Because of Europa's distance from Earth and our inability to send humans there, robots will need to be able to operate without a direct link to human operators.

Therefore, to reduce the cost of ocean science, promote more widespread exploration of the oceans, and enable new and important missions, underwater robots that can perform more complicated tasks without the direct help of human pilots need to be developed. These are called autonomous underwater vehicles (AUVs) because they can complete most tasks autonomously. Not only will these AUVs be able to explore in remote sites like Jupiter's Europa, but they will also help to invert the human-to-robot ratio here on Earth, allowing a single operator to direct and monitor many robots.

Many new capabilities need to be developed in order for AUVs to become more versatile. These include: navigating in unknown regions, selecting scientifically interesting observations, deploying sensors and retrieving data from them, and collecting samples. These capabilities will be developed in various domains, including mechanical engineering, robot path planning and control, sensor design, computer vision processing, artificial intelligence, and more.

My research addresses a critical component of the AUV problem: the ability of an underwater robot to determine its exact position relative to some object in its vicinity. Knowledge of relative position enables relative position control, which is the basis for many important tasks, such as holding station while recording camera images, moving around underwater objects like hydrothermal vents, and manipulating objects with a robot arm. This capability is particularly important for underwater robots because they are always floating and unable to simply “park” or anchor themselves.

Many underwater vehicles already use a variety of sensors to determine their position. A pressure sensor, compass, and inclinometer indicate vehicle depth, heading, and attitude, respectively. Measurements from rate sensors (vehicle acceleration and angular velocity) describe the motion of the robot. Finally, sonar beacons are often used as landmarks to resolve vehicle position. However, all of these common underwater sensors, taken separately or collectively, are incapable of providing complete and accurate robot position relative to an object in its workspace.

Researchers working in other fields of robotics have developed relative positioning algorithms based on computer vision processing. These algorithms track several objects in video camera images acquired at different viewpoints and use these to resolve the relative position between camera and objects. Although these techniques could also be applied underwater, they are generally very complex and not robust enough for the challenging ocean environment.

The goal of my research is to overcome the limitations of these vision-based techniques by merging measurements from acceleration and angular rate sensors into the positioning algorithms. Such a system offers two basic advantages over previous approaches: By including a vision measurement, an accurate relative position estimate is obtained, and by also including measurements from rate sensors, the interpretation of the vision measurement is greatly simplified, and as a result, robust enough for underwater operation. While the benefits of merging measurements from dissimilar sensors are obvious, building algorithms that can accomplish this task is the subject of on-going research.

When integrated with the work of other AUV researchers, the results of this research in underwater robot positioning will enable a new class of AUVs capable of gathering scientific observations without the direct intervention of human operators. For scientists interested in the existence of life on Europa, these new capabilities will enable robotic exploration missions that would not otherwise be possible. For ocean researchers here on Earth, they will provide opportunities to collect more data in more places and at a lower cost, thus enhancing the pace of scientific exploration.