Of Microchips and Men

One of Harvard's most competitive scientific projects lies hidden away on the third floor of Cruft Hall, behind the Science Center, where few non-science concentrators ever venture.

In this out-of-the-way spot, the Harvard Robotics Laboratory--a team of professors, graduate students and undergraduates--races against similar teams across the world to make people out of computer chips and metal parts.

The approximately 30 researchers at the laboratory envision a future with robots acting as reliable errand-runners, able to change batteries in an orbiting satellite or drive to school and pick up the children.

But the visionary team works toward its goal one body part at a time. Right now the staff is concentrating its efforts on perfecting a computer that will imitate human hand-eye coordination. The robot's hand has two fingers and a thumb, each with its own motor, and fingertips that contain high-resolution tactile sensors. Lab experiments test the hand's ability in tasks like lifting and grasping objects.

"It's an obvious challenge if you set out to answer the question 'Can we build intelligent machines?' I think the next step might very well be to put together a mobile robot," said McKay Professor of Applied Mathematics Roger W. Brockett, who coordinates the lab.

A mobile robot is not quite within reach, however, Brockett estimates that it will take between two and 10 years to realize that goal. In the meantime, Harvard will try to complete its hand-eye project before its competitors. "Right now we have our hands full with the hand-eye machine," Brockett said.

The lab works on an annual budget of $750,000, about $250,000 of which goes to fund work on hand-eye coordination in computers. Currently the lab is in its third year of a five-year grant from the National Science Foundation.

And laboratories across the country are working quickly to perfect their own hand-eye machines.

"Multi-fingered hands are particularly competitive at the current time," said Shankar S. Sastry, associate professor of electrical engineering and computer sciences at the University of California at Berkeley.

The Berkeley professor said that improving the hand-eye machines has remained mainly an academic quest, because businesses would rather prefect computer systems they have already invested in than support innovation in the field.

"I kind of share Roger Brockett's point of view that this is the correct thing to do from an academic standpoint even if it isn't what industry is crying for," said Sastry.

The Harvard Robotics Laboratory is also attempting to improve computer vision through work with a binocular machine vision system, which has "eyes" that are small television cameras mounted on motors.

Scientists hope to enable computers to see in three dimensions. In the lab, research teams use 20 to 30 microprocessors--the brains of modern computers--to try to coordinate the computer hand and vision.

"Certainly this binocular head system would be very interesting once we get it going," said James J. Clark, the assistant professor of Electorical Engineering on the Gordon McKay Endowment. "Vision, when you start thinking about it, is a very interesting problem."

The improvement of computer vision could also have business applications.

"Most industrial robots are blind. Just like a blind man, they do something silly when something's been moved out of place from where they expect it. With visual control the robot would be able to see where the object was and adjust," said William A. Woods, McKay Professor of the Practice of Computer Science.

"The next step is to have more flexible robots so that you can manufacture things that have smaller product line runs. Flexible manufacturing could save a lot of inventory costs," said Woods.

Better computer vision could also result in major scientific advances.

"Of course you would like to get a vision system that could act in a totally arbitrary environment," Clarke said, adding that such a computer would be able to function on Mars.

The lab also specializes in a new field of research called Very Large Scale Integration (VLSI), which involves concentrating large numbers of transistors on small silicon chips. More powerful chips mean computers can occupy less space, allowing advances like the installation of more complex systems in cars, and the replacement of racks of telephone wires with small chips.

"Chips are showing up everywhere--microwaves, refrigerators," said Clark. "The more transistors you put on a chip, the more the computer can do."

Because making the chips is an industrial process, it cannot be performed at Harvard. The National Science Foundation supplies the lab with money to have chips made after they have been designed on the Harvard drawing board.

Harvard is also trying to improve the ability of computers to recognize faces. Although lab workers said they are approaching the work from an academic perspective, computers able to recognize individual human faces could make security systems more foolproof.

Scientists trying to recreate human motion in computers refer to something called the "existence proof," Clarke said. If humans can already move in a certain way, researchers need only figure out the mechanics of the motion.

"You know it can be done, it is a challenge to figure out how it's done." Clarke said. "It's an intellectual challenge to reverse-engineer what God has done. It's actually quite amazing how clever the human body is."

Robotics lab staffers work closely as a team, regardless of their individual stature.

"There are people from full professors down to undergrads in [the laboratory], and everybody has the opportunity to get involved in everybody else's projects if they want. There's always something exciting happening," said Michael E. Cohn '89, who works between 10 and 12 hours a week in the lab.

"Professor Brockett is very good about wanting to get lots of kinds of people involved," said Alexander L. Bangs '88, who also works in the lab. "There are robots out there in industry, welding. The thing we're working on is something beyond that."