MIT team finds that the ratio of component atoms is vital to performance.
Lyman Hazelton and colleagues at MIT and NASA's Ames Research Center won't be aboard the shuttle Columbia when it lifts off this month, but a "scientist" they created will be.
That "scientist," actually a system known as the Astronaut Science Advisor (ASA), "is the first computerized scientific assistant based on artificial intelligence to go into space," said Dr. Hazelton, a research scientist at the Man-Vehicle Laboratory in the Center for Space Research and chief scientist for the ASA project (a collaboration between MIT and the Ames Research Center).
Professor Laurence R. Young of the Department of Aeronautics and Astronautics is principal investigator for the project (Dr. Young originated the idea for the ASA during a sabbatical at the Ames Research Center), and Associate Professor Peter Szolovits in the Laboratory for Computer Science is a co-investigator for the project.
The ASA could lead to artificial-intelligence systems in space and on Earth capable of running-and modifying-experiments by themselves. Such systems would have particular relevance for research aboard unmanned space missions and at infrequently visited locales on Earth. The ASA, however, will be controlled by astronauts.
The Columbia mission, Spacelab Life Sciences 2, will mark the ASA's first time in space (liftoff is scheduled for October 14). While there, the system will aid astronauts conducting the rotating-dome experiment (see main story) by suggesting alternative experimental protocols when appropriate, flagging interesting data that the astronauts might want to explore further, and more. Such suggestions, the scientists who developed the ASA hope, should make the most efficient use of the astronauts' time and lead to more scientifically valuable data.
Currently "it is very difficult to do research in space," Dr. Hazelton said, "because most of the time the scientists behind the experiments are not on the flights." The astronauts who conduct the experiments are highly trained, but they cannot be expected-and often don't have the time-to modify an ongoing experiment, for example, should incoming data warrant a change.
As a result, Dr. Hazelton said, the ground-based scientists "often have to propose another flight of the experiment" to incorporate modifications, "which is both expensive and exasperatingly slow."
The ASA "is a first step toward circumventing [such] limitations," Dr. Hazelton said. At the request of the astronaut operating the rotating dome, the ASA will review the data collected and modify further runs of the experiment based on those data. In addition, should some of the time originally allocated for the experiment get cut, the ASA can "suggest a streamlined protocol to get only the most essential data," he said.
Similarly, if extra time becomes available the ASA can generate additional experimental protocols. "Any member of the crew can create a new, scientifically appropriate experimental protocol for the rotating dome experiment in less than two minutes," Dr. Hazelton said. "We've tried to make it as convenient as possible for the astronauts to conduct extra runs."
The ASA, which operates on an Apple Macintosh PowerBook computer using a combination of software available commercially and developed by MIT and NASA, also contains a trouble-shooting and repair module that will aid astronauts should any part of the experimental apparatus malfunction. The module comes complete with the ability to decide whether a given problem is worth repairing. Explained Dr. Hazelton: "If you're five minutes into the experiment and a light burns out, it's probably worth replacing the bulb. But if that light burns out toward the end of the experiment, it's probably better to let it go. The machine considers these different weightings."
If the astronaut does decide to fix the problem, the module will give step-by-step instructions-with labeled drawings-and indicate the location of tools.
Finally, as Dr. Hazelton is proud of noting, the ASA is comparatively inexpensive and took much less time to prepare for flight than other experiments. This was because the scientists got NASA to hold the ASA to less rigorous standards by deeming it a "non-critical" addition to the payload. As a result, they didn't have to "effectively guarantee that it would work," Dr. Hazelton said, though they did have to show that it wouldn't affect data collection for the dome experiment or hurt the astronauts or any of the other experiments. (The ASA computer's interface to the rotating dome, however, was held to normal Class C standards.)
"So instead of having to spend one to two million dollars to qualify the ASA by Class C standards, we spent $200,000 to qualify it by Class D standards," Dr. Hazelton said. "As far as I know, this is the only experiment that's been done this way."
A version of this article appeared in the October 6, 1993 issue of MIT Tech Talk (Volume 38, Number 9).