Known as Remote Agent, the software has been operating NASA's Deep Space 1 mission and its futuristic ion engine since 11 a.m. Pacific Daylight Time today, May 17. The question: Can a spacecraft function entirely on its own nearly 120 million kilometers (75 million miles) from Earth, without detailed instructions from the ground?

The public is invited to follow this ambitious 48-hour test as it continues to unfold today and tomorrow, through a detailed web page and e-mail alerts triggered by actual events on Deep Space 1.

"While we watch over its shoulder electronically, we are giving Remote Agent the responsibility to monitor Deep Space 1's activities and position in space, including any engine firings it needs to keep on course," said Dr. Pandu Nayak, deputy manager of Remote Agent development at NASA's Ames Research Center, Moffett Field, CA. "We are also challenging Remote Agent with some 'unexpected problems' to see how well it reacts, and to determine whether it can get the mission back on track without human intervention."

"Remote Agent can create and carry out its own plans to achieve the mission goals that we give it," said Dr. Doug Bernard, Remote Agent manager at NASA's Jet Propulsion Laboratory, Pasadena, CA. "This technology could allow us to pursue solar system exploration missions that only a few years ago would have been considered too elaborate, too costly or excessively dependent on teams of Earth-bound controllers."

The Remote Agent software package features three components: the Planner/Scheduler, the Executive and one called Livingstone.

The Planner takes general goals and determines detailed activities needed to achieve the goals. The test includes asking the Planner to achieve broad goals such as, "Find your position, and fire your ion engine whenever practical." If a hardware problem develops that prevents execution of the plan, the Planner makes a new plan, taking into account degraded capabilities.

The Executive interprets the plans and adds more detail to them, then issues commands to the flight software, coordinating the three parts of Remote Agent. Some commands turn the spacecraft to point in a different direction. Other commands ask the onboard camera to take pictures of asteroids and stars for navigation purposes.

Livingstone acts like a doctor, monitoring the spacecraft's health. If something goes wrong, Livingstone tells the Executive there is a problem. The Executive consults the "doctor" for simple procedures that may quickly remedy the problem. For example, if the camera does not respond, a quick fix is to turn the camera off and then on again. If this does not work, the Executive asks the Planner for a new plan that still achieves mission goals. If the problem is too serious, the software gives up and waits for help from Earth.

Specific tests include simulation of hardware problems, such as a malfunctioning spacecraft thruster. This should prompt the software to diagnose the cause of the apparent problem and take corrective action. Remote Agent is designed to detect and recover from a set of real subsystem failures in the unlikely event that an actual failure should occur on Deep Space 1 during the remainder of the experiment.

Launched October 24, 1998, Deep Space 1 is validating 12 new technologies, including Remote Agent, so that they can be confidently used on science missions of the 21st century. Its ion propulsion system has now completed more than 73 days of thrusting, most of that time under the control of the spacecraft's autonomous navigation system. The Deep Space 1 team expects that testing of all technologies will be complete by early June, with the exception of navigation system tests scheduled during an encounter with asteroid 1992 KD in late July.

The Remote Agent software was developed in collaboration between NASA Ames and JPL.