All systems go for NASA tech

NASA's Deep Space 1 mission, which formally ended in December, began as a test bed for risky new technologies and ended in the startling success of artificial intelligence systems that will make possible missions that could not have been considered before.

The AI-based expert systems were used to control a number of critical processes. After Deep Space 1, expert systems — software designed to mimic human decision-making — will take over many of the routine tasks of future missions, as well as some of the more complex procedures, said Ed Reidel, the mission's navigation team chief.

"The AI used on Deep Space 1 is probably the most capable of the space software technology that we've put out there, as far as an automated capability working over long periods," he said. "Deep Space 1 was a milestone for the use of these technologies."

Reidel's team was directly responsible for autonomous navigation (AutoNav), the on-board AI system that generated the most buzz. It was designed to guide the spacecraft over long distances with pinpoint accuracy and to help in various complicated maneuvers, all with minimal interaction with ground controllers.

Given the distance the craft would be from Earth, radio signals from controllers would take too long to get to it to allow for "joystick" navigation from the ground. AutoNav used the positions of some 250 asteroids and compared them with those of 250,000 stars stored in the on-board computer to check the craft's position and make corrections.

Running on a 20 MHz radiation- hardened version of IBM Corp.'s RS/6000 chip, with just 4M of memory, Auto- Nav guided the craft to within 15 miles of an asteroid called Braille in a flyby at the end of the primary mission in 1999.

Unfortunately, the craft's star- tracking system — which took pictures of stars and locked on to one as a reference point that AutoNav used in its calculations — failed just after the Braille flyby. AutoNav could not then be used to direct the craft but was used later, after reprogramming from the ground, to help analyze pictures taken by the on-board camera so that the craft could once again locate a star.

"I think AutoNav is a wonderful technology," said Marc Rayman, Deep Space 1's chief mission engineer and deputy mission manager. "And we all appreciated it even more after the star tracker failed. We didn't have AutoNav controlling the ion propulsion system anymore and had to do that from the ground, which involved much more intensive work on our part. That clearly demonstrated just how AutoNav was able to lift the human burden of flying spacecraft."

Other expert system technology tested on Deep Space 1 included Remote Agent, a system used to generate and execute plans on board for such things as camera imaging, and a beacon monitor system, which regularly scanned the craft's health and alerted ground controllers by sounding one of four tones for the degree of urgency of any problems found. Although considered less important than AutoNav, both systems also tested perfectly, Rayman said.

Robinson is a freelance journalist based in Portland, Ore. He can be reached at hullite@mindspring.com.