Popular Mechanics - April 1993
(page 130-131
I wasn't expecting much when a derailed train left me with a day to kill in Salt Lake City, Utah. But science reporting is often as much a matter of serendipity as advance planning, so I decided to pass up the chance to see the Mormon Tabernacle and pay a surprise visit to the University of Utah. As it happened, my visit coincided with some big news from the university's robotics researchers. Score another one for serendipity. Dr. Stephen C. Jacobsen, rootics guru and head of the College of Engineering's Center for Engineering Design, has one of the most hyperorganized offices I've ever seen. Cabinets full of meticulously color coded files line the walls, and his phone projects toward him on an articulated stalk. When I walk in, he seems entranced with the images on a large computer monitor. You get the sense that this is a guy whose time is so valuable that everything around him must be structured perfectly to prevent a single moment's thought or movement from going to waste. And indeed, his efforts have been instrumental in advances ranging from the Utah Artificial Arm, which now gives renewed dexterity to hundreds of amputees, to manipulators for secretive Defense Department projects and beguilling animatronics figures at theme parks. After a pause, he swivels his chair slowly to face me like some James Bond supervillain. But he is cordial and speaks with precisely measured enthusiasm. At the moment, he's particularly excited about a new class of tiny sensors that will help robots and other intelligent machines monitor their own movements. "That's the biggest thing out of the lab," he says flatly.
At first, it doesn't seem like such a big deal, but with a little explanation,
the implications grow clearer. "Something that has permeated this place for
years is a sadness that the strategies used by biology can't by used by manmade
machines," Jacobsen says. "That is, using large numbers of sensors that acquire
precision from statistics, not measurement, and work in groups and networks
to interpret machines such that they can function like organisms. The option
of building machines that work in the same strategic ways as biological machines
has been unavailable because of the cost of the sensors, and what this is
going to do is change the philosophy of design for machines."
Dr. Melvin W. Siegel of the Robotics Institute at Carnegie Mellon University agrees that the sensors have great promise. It stems, he says, from their ability to provide something called proprioception, a sexth sense which we all take for granted that tells us what our own parts are up to at any given moment. "That's what robots don't have," says Siegel. "They're running blind and the existence of small, lightweight, reliablee sensors will let us work with machines that know where their own parts are. You know how important that is to you." One example is an unimpressive looking little device called a Rotary Displacement Transducer (RDT). Consisting of a cylinder about the size of three stacked dimes with a shaft protruding frome one end, its purpose is to provide data on the rotation of one part with respect to another. It does this in a unique but simple way. Inside the cylinder, a wheel etched with electric emitters in an elaborate pattern known as a Gray code totates on the end of the shaft. The coded electric field it produces is read by a gold plated microchip ring fixed to the bottom of the cylinder. This simplicity yields a $50 unit the size of a marble that can replace limon-size devices costing as much as $1400. The data it produces is extremely precise, indicating movements as small as .018". Perhaps even more important, this data comes in the form of multiplexed digital signals, so information from up to 128 sensors can travel along the same set of three wires.
The principles behind the RDT have spawned a whole class of sensors offering
similar advantages. Taken together, these should dramatically reduce the complexity
and cost of a wide range of machinery, which earlier relied on analog sensors,
requiring great bundles of wiring for communication. Botton line, says Jacobsen:
"It'll probably take 30% off the cost of a robot."
And the dividends don't stop there. "The encoder business is over a billion dollar market-everywhere from washing machines to airplanes," he says, adding that major automakers have already expressed interest. So how come nobody came up with anything like this before? Probably because microchips had always been thought of as tools for processing data, rather than for generating it. For all their electronic brainpower. they were believed too fragile to interact directly with the real world. meaning that you always had to have two separate devices-one to do the sensing, and one to make sense of the sensor. "Nobody's ever had the stupidity or whatever to say we're going to take a device and put it right on the chip and we're going to treat the chip so it can stand it," says Jacobsen. But with the Rotary Displacement Transducer and its brethren, Jacobsen's group has done exactly that. "The chip, instead of extracting information from wires and from optical processors and strain gages, is looking directly at the physical process and extracting information. That means lower cost, higher bandwidth and smaller size," he says Before my meeting with Jacobsen, I had a chance to look around the CED with Todd Johnson, the lab's senior electrical engineer. Thinking back on what he showed me, I could see the potential of the new developments. Although designed before the latest sensing technology was available, the robots populating the place already had hauntingly lifelike powers. Yet, a reverence for the way nature does things was a theme repeated throughout the tour and an obvious source of inspiration for much of the design work, "look at the power it takes to move an artificial arm," said Johnson at one point. "Compare that to biology, where a Snickers bar is good for 2 or 3 hors on the ski slope. Then you begin to realize, well, first of all, you have to have a lot of respect for what nature has achieved. Then you have to go see how much of it you can understand and perhaps use."
SIXTH SENSORS