Computerized valet kept tabs on astronaut Glenn
- By Richard W. Walker
- Jan 25, 1999
When John Glenn donned a wired vest and head gear for a sleep experiment on his recent
space shuttle mission, he had a helper, a kind of computerized valet to make sure
everything was properly connected and activated.
Glenns helper was an artificial intelligence system called the Principal
Investigator in a Box, developed by researchers at the Massachusetts Institute of
Technology and NASAs Ames Research Center at Moffett Field, Calif. PI in a Box acts
as the eyes and ears of the research scientist who cannot be on the shuttle.
For four nights during the flight, Glenn and fellow astronaut Chiaki Mukai wore
instruments to measure vital signs related to sleep and respiration.
Once Glenn and Mukai were rigged up, astronauts Stephen Robinson and Scott Parazynski
connected the apparatus to a digital sleep recorder that was in turn plugged into an IBM
ThinkPad notebook computer running the PI in a Box application. The system displayed and
analyzed signals for 15 parameters, such as brain wave, eye movement, muscle activity and
For the user, PI in a Box works a bit like a traffic light. It first checks that all
signals are present and then assesses them for acceptable range. If a signal meets the
test, a green light appears on the notebooks screen next to a chart describing the
signal. An amber light indicates the system is still analyzing the data.
If a signal is absent or of anomalous quality, a red light flashes on. The userin
this case a fellow astronaut acting as proxy for the scientist on the groundclicks
on the red light, prompting a dialog box containing troubleshooting procedures.
When all signals are green, the experiment is a go. The user shuts down the PI in a Box
and disconnects the ThinkPad from the sleep recorder.
PI in a Box is the brainchild of Laurence Young, Apollo professor of astronautics at
MIT and director of NASAs National Space Biomedical Research Institute.
Young in the early 1980s began to think about systems that would apply a
researchers reasoning power to remote experiments in space. When the 1986 Challenger
tragedy brought NASA missions and his own space shuttle experiments to a halt, he decided
to take advantage of the delay and develop his idea.
I wondered how I could use this new technology of artificial intelligence to do
my job better, training astronauts to be my eyes and ears in space, Young said.
It occurred to me that if there was some way I could, short of being up in space
myself, give my reasoning power by means of a computer [link] to astronauts who were doing
my experiment, I could vastly increase productivity in space.
He took a six-month sabbatical at NASAs Ames Research Center to study artificial
intelligence. Young and a group of researchers at Ames and MIT later began to develop and
test the PI in a Box. They sent an early version into space on a 1993 mission to support
an experiment by MIT scientists on human equilibrium.
PI in a Box uses an expert shell systema reasoning structure without any
knowledge in itcalled the C Language Integrated Production System, or CLIPS,
developed in the mid-1980s at the Johnson Space Center in Houston and now maintained as
public domain software by the main program authors, who no longer work for NASA.
To provide the knowledge the expert system needed to evaluate the sleep
instrumentations setup, Youngs team conducted a series of interviews with the
group of scientists working on the experiment, headed by Charles Czeisler of Harvard
Medical School and Brigham and Womens Hospital in Boston.
The system currently runs on an older ThinkPad with a 486 microprocessor, Young said.
One of the frustrations is that we have lots of better computers around, but in the
NASA world, youre forced to fly what have already been flight-qualified and
certified, he said. This was the computer that NASA had in its stable of
previously flown and qualified computers.
PI in a Box was first used for sleep experiments in space last April on the Neurolab
mission, a two-week life sciences research flight. During setup, the system alerted
astronauts to anomalies in signals more than 100 times.