Robotics

The independent study project undertaken by Andrei Furtuna '05 should appeal to fans of Isaac Asimov's science fiction. Under the supervision of Rebecca Thomas, associate professor of computer science, Furtuna constructed and programmed several small robots. He employed Agent Oriented Programming (AOP), an experimental software paradigm, to render the robots capable of working cooperatively to perform tasks, such as picking up and transporting objects. Infrared communication, controlled via a movable shutter, is used to coordinate the robots' actions. AOP is a relatively new and underexplored area in science.

Thomas, who is a recognized expert in Agent Oriented Programming, describes Furtuna's work as "a novel application of the AOP paradigm, which has been used previously to program more complex software systems." She also notes, "We never envisioned using AOP on machines like these robots. This approach will help us learn about the limitations of the AOP approach to programming. If the experiment succeeds, it has broad implications. AOP is a looser, vaguer type of instruction than we normally give to a computer. The result is a robot that's more adaptable to external circumstances."

Thomas wrote her Ph.D. dissertation on various artificial intelligence problems that the AOP approach might solve. The field that she is exploring (and that she has introduced to Furtuna) seeks solutions to those real-world external factors that hamper the use of robots in situations where robots could play an essential role. That is, if robots could cooperate. AOP, based on an "if. . . then" approach to programming, offers options and defaults as circumstances arise and change.

A hypothetical example of one such circumstance might be a mission to Mars that involved, perhaps, as many as 50 robots. Like many robots currently in use, each would have a dedicated task, such as sampling rocks or transmitting data back to a space station on earth. If three robots were dedicated to transmitting data and eight were detailed to taking rock samples, what would happen if one of the three transmitting robots were to malfunction? Could the programs in the remaining two transmitting robots and in the eight sampling robots be written in such a way so that the robots could adjust to the situation and find a new pattern of cooperation, just as humans could?

In Thomas's popular Introduction to Robotics course, these kinds of scenarios are examined. Each student in the course receives a LEGO Mindstorms kit—modern kits are a far cry from the early, unsophisticated LEGO sets. Like Furtuna, the students not only build a working robot with the kit, they learn how a central processing unit works, the physics behind the use of motors and gears, and how to write a computer program. By coping with many of the same hurdles that face high-tech engineers, the students also familiarize themselves with artificial intelligence's current strengths and limits. "There are common programming problems, no matter what kind of robot you are talking about, whether it starts as a LEGO kit or as an industrial model," says Thomas.

For instance, a robot cannot see in the way a human can. Instead, it can use lasers, shot off at timed intervals, to determine how close something may be. However, that information has a built-in margin of error. Even something as seemingly simple as directing a robot to move forward will be affected by any change in the surface on which the robot is operating. Thomas notes, for example, that many of her students have programmed robots that function perfectly on a bare floor but are rendered immobile by a pile carpet. These problems constitute an enticing challenge for the Bardians.

With course offerings like Introduction to Robotics, Thomas and several other professors in Bard's Division of Science, Mathematics, and Computing have taken a creative approach to teaching those students who do not plan to concentrate in the sciences. The professors, feeling that traditional, basic introductory courses rarely touch on a field's true intellectual challenges, have chosen to plunge students directly into fascinating issues.

Bard's new requirement structure has spurred the professors' efforts. Under the recently revised system, students must earn credit in quantitative thinking, as well as in the natural sciences. While students may earn quantitative credit by taking advanced courses in physics or chemistry, many nonscience students opt for a semester of computer science or mathematics. "There's a concerted effort at Bard to integrate the sciences into the whole curriculum," says assistant professor of computer science Sven Anderson, who teaches a computer simulation course designed for nonscience students. "Bard wants science to be relevant to all its students. Part of a liberal arts education is realizing that there are interesting ideas in any field."

To foster that relevance and interest, Anderson is constructing a course in cognitive science, an offering usually available only as a survey course in large universities. The Bard course, scheduled for inclusion in the spring 2005 curriculum, will constitute, says Anderson, "a discussion and exploration of what underlies intelligence and thinking." In that exploration, the course will not limit itself to considering only the human brain. Instead, it will examine the ability to learn, understand, or reason as it exists in, for example, electronic or mechanical forms. The course will be interdisciplinary (drawing on linguistics, computer science, philosophy, psychology, mathematics, biology, chemistry, and other fields), will involve a number of faculty members, and will include lab work. Of the material that will be covered, Anderson says, "I expect it to be congenial to nonscientists, but compelling enough for scientists."

Several other courses in Bard's Division of Science, Mathematics, and Computing have already taken an innovative tack. Lauren Rose, associate professor of mathematics, has received excellent student response to her Explorations in Number Theory course. In the course, students work in small groups, often arriving at theorems or proofs that were, relatively recently, among the field's biggest brainteasers. Other exercises have hands-on applications. One, for example, involves learning how to encrypt communications in a number-based code similar to the one used to secure the amazon.com website. Students then send encrypted messages to one another. "I wanted to show students how to think like mathematicians," Rose says. "Mathematicians don't just compute. They create and discover."

Bard students also have the option of approaching mathematics by delving into its history. Jeffrey A. Suzuki, visiting assistant professor of mathematics, teaches Mathematics of the Premodern Era, a course that traces the mathematical methods used by various civilizations, such as ancient Egypt, Mesopotamia, China, India, and Islamic and early Western cultures. By developing proficiency in the mathematical systems used during different historical eras, students acquire an extensive vocabulary of alternative ways to derive valid solutions to problems in numeration, computation, geometry, root extraction, algebra, multilinear systems, and third- and fourth-degree equations.

In another course, Simulating Reality (taught by Anderson), the entire world becomes fodder for students to examine how computers create models for everything, from corporate design to children's games. Students also examine real-world problems, such as the ways in which models for disease control can help stem epidemics, and each student develops a simulation project of his or her own. Recently, one student with an interest in dance explored whether a program that simulated human body movement could record choreography as well as labanotation does.

Once nonscience students discover that computer science and math offer plenty of room for creative speculation, many go on to take a second or third course in the division. Thomas lauds students' willingness to explore ever-expanding areas, such as artificial intelligence. "Where we are now with robots is comparable to where we were with personal computers in the 1970s," she explains. "You used to type letters, or equations, in green letters on a black screen—a Windows-based system was unimaginable. In 10 or 20 years we will see similar improvements in robotics." Although, at the moment, robots' uses are limited by their processing units' inability to assimilate information in real time, eventually, increased capability will dispense with those limitations. In fact, considering that the new Gabrielle H. Reem and Herbert J. Kayden Center for Science and Computation (see page 6) will have a computer lab equipped with more powerful robots than have been used in the curriculum so far, Thomas has a fantasy that, perhaps in 10 years, robots roaming the science building will be capable of running errands on command.

—Hanna Rubin

"There's a concerted effort at Bard to integrate the sciences into the whole
curriculum," says assistant professor of computer science Sven Anderson.

"I wanted to show students how to think like mathematicians," Rose says.
"Mathematicians don't just compute. They create and discover."