NASA scientists recently radioed artificial intelligence (AI) software successfully to a satellite and tested the software's ability to find and analyze errors in the spacecraft's systems. Normally, troubleshooting is done on the ground.

The AI software, Livingstone Version 2 (LV2), automatically detects and diagnoses simulated failures in the NASA Earth Observing One (E0-1) satellite's instruments and systems. E0-1, launched in November 2000, is a flying test bed for new technologies and techniques intended to boost safety, and to reduce costs and development times.

"This is the kind of technique and technology NASA needs to support exploration of the Earth, moon, Mars and beyond in the 21st century," said Ghassem Asrar, deputy associate administrator for NASA's Science Mission Directorate. "This software grants us the ability to troubleshoot the robotic systems required to handle increasingly complex tasks of exploration, while they are millions of miles and perhaps light years away from Earth," he said.

"Livingstone gives us a chance to recover from errors, protect our investments in space and continue to achieve our mission goals," said Sandra Hayden, the Livingstone E0-1 experiment principal investigator. She is a scientist at NASA Ames Research Center, located in California's Silicon Valley. "It is critical to ensure spacecraft systems behave as designers intended, and to accurately diagnose potential problems," Hayden said.

Tests of the LV2 computer program are taking place while another software application controls E0-1. NASA's Jet Propulsion Laboratory (JPL), Pasadena, Calif., is conducting the Autonomous Sciencecraft Experiment (ASE), which is controlling EO-1. NASA's Goddard Space Flight Center (GSFC), Greenbelt, Md., manages E0-1 operations.

LV2 is monitoring the ASE software, as it autonomously runs the satellite's imaging systems. If the EO-1 does not respond properly to ASE control, then LV2 detects the error, makes a diagnosis and radios its analysis to mission control at Goddard. LV2 will decide the best way future missions with subsystem failures can continue and still achieve goals.

LV2 works by comparing a computerized model of how the spacecraft's systems and software should perform against actual performance. If the spacecraft's behavior differs from the model, then the LV2 'reasoner' looks for the root cause of this difference and gives flight controllers several suggestions of what might have gone wrong.

Scientists designed the LV2 software 'reasoner' independent of the computerized model of the spacecraft systems. To use it on another satellite, an engineer can apply much of the system model created for a previous spacecraft. The model may be customized for the small set of differences required to interface with the 'reasoner.' This is because many spacecraft use common parts, such as valves, switches and sensors, which have the same or similar behavior patterns. All LV2 E0-1 experiments have been successful, with the rest scheduled over the next few weeks.

Livingstone's developers feel the software diagnosis tool could be used to find errors in robots or rovers exploring Mars or other planetary bodies. Engineers state that when human beings venture deeper into space, crews will need automatic tools like Livingstone software to identify spacecraft problems early and make prompt repairs.

"In a future, long-duration human mission to Mars, frequent and extensive spacecraft maintenance operations or overhauls will not be an option," explained Serdar Uckun, technical lead for integrated systems health management at NASA Ames.

Scientists predict Livingstone software and its descendents will find widespread use at NASA, in the aerospace industry and in other enterprises as equipment and software intricacy increases.