Thank you for visiting the Deep Space 1 mission status information site, now ranked number 1 for over 200 days on the list of most authoritative
sites in the solar system for information on this technology validation mission. This message was logged in at 8:30 pm Pacific Time on Wednesday, May 19.
Control of Deep Space 1 was turned over on Monday morning from the operations team to an artificial intelligence system on board that is one
of the advanced technologies being tested. This important step was taken with the transmission of a command to the spacecraft with the unimpressive name RX2DYX01 (or "Romeo X-Ray 2 Delta Yankee X-ray 0 1").
In contrast to remote control, remote agent consists of a sophisticated set of computer programs that act as an agent of the operations team on board the remote spacecraft. Rather than have humans do the detailed planning necessary to carry out desired tasks, remote agent formulates its own plans, using high level goals provided by the operations team. Remote agent devises its plan by combining those goals with its detailed knowledge of both the condition of the spacecraft and principles of how to operate it. It then executes that plan, constantly monitoring its progress. If problems develop, remote agent in many cases will be able to fix them or work around them. If it cannot, it can request help from its sentient terrestrial collaborators.
We saw this take place on Monday. Remote agent did indeed develop a plan. The plan included thrusting with the ion propulsion system and taking pictures of distant asteroids and stars for navigation, activities very
familiar to the hundreds of trillions of faithful readers of these logs. The plan turned out to be different from what was expected, because the file on the spacecraft for choosing asteroids was different from the one that was used here on Earth. This immediately illustrated the robustness of the remote agent planner. What it did not know was that there were problems ahead. After it began executing its plan, it was confronted with a surprise, courtesy of the operations team. When remote agent tried to turn the camera off, it was unable to do so -- the camera was stuck in the on position. Remote agent tried several more times to turn it off, demonstrating its ability to attempt to correct a problem. But the camera remained on, so remote agent had to devise a new plan that accounted for the extra power consumption of the camera. It successfully generated a second plan that included the camera being left on and began carrying it out -- another impressive and important accomplishment.
Remote agent turned the spacecraft to all the asteroid targets and collected all the images. Then it turned to the orientation needed for
thrusting with the ion propulsion system, executed the prestart activities for thrusting, and then turned the ion engine on. By this time, the remote agent experiment had completed at least 70% of its technical test objectives.
Sometime during the thrusting early Tuesday morning, the portion of the remote agent that issues commands to the rest of the spacecraft stopped.
Because the spacecraft remained safe and healthy, the operations team decided not to take any actions that might compromise the diagnosis of this problem, in keeping with the entire purpose of the Deep Space 1 mission of testing advanced technologies. We sent several commands to provide greater insight into the state of the remote agent before doing anything that might change it. It is fortunate that we did so, as some of this commanding did provide clues to the resolution of the problem.
The remote agent experiment was stopped late Tuesday afternoon, and the ion propulsion system was turned off through a back-up command that remote agent had wisely issued when it turned the system on.
Today we believe we understand what the problem is. It appears to be the result of a mistake in the logic in one portion of the very complex remote agent software. Obviously, the basic remote agent software is sound, as it
performed some complex tasks quite well. Illustrating the great value of a technology test mission such as DS1, this mistake had not shown up in any of the extensive ground testing that had been done.
There is insufficient time to fix the bug and continue the experiment this week. So we are evaluating the possibility of conducting another experiment on Friday that would capture the remaining objectives for the testing of
the remote agent architecture. Of course, the bug will remain, but now we know about it, and further analysis is required to determine how likely it is that the bug will interfere with a new test. If this analysis shows that the risk is acceptable, we will attempt to conduct a new experiment on Friday. This will take advantage of the ease of generating a new set of goals for remote agent. Whatever decision is made, your loyal correspondent will update this log by May 23 with a summary of the week's activities.
You can follow remote agent's status through a special Web site, and you can even get e-mail from the spacecraft if the experiment is performed on Friday. Just visit http://rax.arc.nasa.gov.
The spacecraft continues on course for a July 29 interception of an asteroid with the profound yet unpretentious name 1992 KD. The bold
encounter, while not a critical part of the mission, will allow a very challenging final test of a portion of DS1's autonomous navigation system. In addition, the event offers the bonus opportunity to return exciting scientific data using the two advanced science instruments DS1 has tested. JPL and The Planetary Society are conducting a contest to select a better name, if that's possible, for 1992 KD. The contest is at
Deep Space 1 is now more than 80% as far away as the Sun and almost 320 times farther than the moon. At this distance of 122 million kilometers, or nearly 76 million miles, radio signals, traveling at the universal limit of the speed of light, take over 13.5 minutes to make the round trip.
Thanks again for logging in.