Brooks, Rodney

Brooks, Rodney

(1954–  ) Australian, American

Roboticist

Rodney Brooks’s ideas about robots have found their way into everything from vacuum cleaners to Martian rovers. Today, as director of the Artificial Intelligence Labora-tory at the Massachusetts Institute of Technology, Brooks has extended his exploration of robot behavior into new approaches to artificial intelligence.

Brooks was born in Adelaide, Australia, in 1954. As a boy he was fascinated with computers, but it was still the mainframe era, and he had no access to them. Brooks decided to build his own logic circuits from discarded electronics modules from the defense laboratory where his father worked. Brooks also came across a book by Grey Walter, inventor of the “cybernetic tortoise” in the late 1940s. He tried to build his own and came up with “Nor-man,” a robot that could track light sources while avoiding obstacles. In 1968, when young Brooks saw the movie 2001: A Space Odyssey, he was fascinated by the artificial intel-ligence of its most tragic character, the computer HAL 9000 (see artificial intelligence and robotics).

Brooks majored in mathematics at Flinders University in South Australia, where he designed a computer language and development system for artificial intelligence projects. He also explored various AI applications such as theorem solving, language processing, and games. He was then able to go to Stanford University in Palo Alto, California, in 1977 as a research assistant.

While working for his Ph.D. in computer science, awarded in 1981, Brooks met John McCarthy, one of the “elder statesmen” of AI in the Stanford Artificial Intelli-gence Lab (SAIL). He also joined in the innovative projects being conducted by researchers such as Hans Moravec, who were revamping the rolling robot called the Stanford Cart and teaching it to navigate around obstacles.

In 1984 Brooks moved to the Massachusetts Institute of Technology. For his Ph.D. research project, Brooks and his fellow graduate students equipped a robot with a ring of sonars (adopted from a camera rangefinder) plus two cameras. The cylindrical robot was about the size of R2D2 and was connected by cable to a minicomputer. However, the calculations needed to enable a robot to identify objects as they appear at different angles were so intensive that the robot could take hours to find its way across a room.

Brooks decided to take a lesson from biological evolu-tion. He realized that as organisms evolved into more com-plex forms, they could not start from scratch each time they added new features. Rather, new connections (and ways of processing them) would be added to the existing structure. For his next robot, called Allen, Brooks built three “layers” of circuits that would control the machine’s behavior. The simplest layer was for avoiding obstacles: If a sonar signal said that something was too close, the robot would change direction to avoid a collision. The next layer generated a random path so the robot could “explore” its surroundings freely. Finally, the third layer was programmed to identify specified sorts of “interesting” objects. If it found one, the robot would head in that direction.

Each of these layers or behaviors was much simpler than the complex calculations and mapping done by a tradi-tional AI robot. Nevertheless, the layers worked together in interesting ways. The result would be that the robot could explore a room, avoiding both fixed and moving obstacles, and appear to “purposefully” search for things.

In the late 1980s, working with Grinell More and a new researcher, Colin Angle, Brooks built an insectlike robot called Genghis. Unlike Allen’s three layers of behav-ior, Genghis had 51 separate, simultaneously running com-puter programs. These programs, called “augmented finite state machines,” each kept track of a particular state or condition, such as the position of one of the six legs. It is the interaction of these small programs that creates the robot’s ability to scramble around while keeping its balance. Finally, three special programs looked for signals from the infrared sensors, locked onto any source found, and walked in its direction.

Brooks’s new layered architecture for “embodied” robots offered new possibilities for autonomous robot explorers. Brooks’s 1989 paper, “Fast, Cheap, and Out of Control: A Robot Invasion of the Solar System,” envisaged flocks of tiny robot rovers spreading across the Martian surface, exploring areas too risky when one has only one or two very expensive robots. The design of the Sojourner Mars rover and its successors, Spirit and Opportunity, would par-tially embody the design principles developed by Brooks and his colleagues.

In the early 1990s Brooks and his colleagues began designing Cog, a robot that would embody human eye movement and other behaviors. Cog’s eyes are mounted on gimbals so they can easily turn to track objects, aided by the movement of the robot’s head and neck (it has no legs). Cog also has “ears”—microphones that can help it find the source of a sound. The quest for more humanlike robots continued in the late 1990s with the development of Kismet, a robot that includes dynamically changing “emo-tions.” Brooks’s student Cynthia Breazeal would build her own research career on Kismet and what she calls “sociable robots” (see Breazeal, Cynthia).

By 1990, Brooks wanted to apply his ideas of behavior-based robotics to building marketable robots that could perform basic but useful tasks, and he enlisted two of his most innovative and hard-working students, Colin Angle and Helen Greiner (see iRobot Corporation). The com-pany is best known for the Roomba robotic vacuum cleaner. Brooks remains the company’s chief technical officer.

Meanwhile Brooks has an assured place as one of the key innovators in modern robotics research. He is a Found-ing Fellow of the American Association for Artificial Intel-ligence and a Fellow of the American Association for the Advancement of Science. Brooks received the 1991 Com-puters and Thought Award of the International Joint Con-ference on Artificial Intelligence. He has participated in numerous distinguished lecture series and has served as an editor for many important journals in the field, including the International Journal of Computer Vision.