Saving Fermi: NASA’s system for avoiding collisions with space junk Finding your space telescope has a dead Russian spy satellite coming at it.

In late March last year, the people operating the Fermi Gamma-ray Space Telescope got a bit of a scare. Their hardware was one week away from a close encounter with a defunct Russian spy satellite. A week might seem like short notice for one-of-a-kind hardware like Fermi, but in some ways the team was lucky to have any warning at all. Prior to 2007, NASA didn't even have a policy in place to identify threats to unmanned hardware.

That has now changed. Thanks to cooperation between the military and the Goddard Space Flight Center, everything that can possibly get out of the way of space junk is regularly tracked for potential collision risks. That system is what alerted Fermi's controllers to the danger and allowed them to use on-board thrusters for something they were never designed to do: move the satellite safely out of the way.

Tracking threats

The technical term for the risk identification process is "conjunction assessment," or CA. This process involves taking the latest tracking information, calculating orbits, and figuring out whether two objects are likely to be in the same place at the same time. This may seem like an easy problem—after all, it's just math—but there are a lot of potential complications. One is simply the ever-expanding catalog of junk in the relevant orbits (primarily low-Earth and geostationary), some of it caused by collisions between two existing pieces of junk. Another is the fact that these orbits aren't stable. The atmosphere may be sparse at these altitudes, but it's still there, and it creates varying amounts of drag on things in orbit.

Finally, tracking junk is not a matter of knowing precisely where everything is at all times. The measurements come with uncertainties and aren't constantly updated, so these uncertainties can grow with time. At best, you can typically only calculate the probability of a collision rather than getting a binary yes-or-no answer.

The Goddard Space Flight Center started this sort of monitoring in 2004, but things were formalized in 2007 when NASA issued its first formal policy. If the hardware was in or near low-Earth or geosynchronous orbit and it had the capability to maneuver in orbit, it needed to have conjunction assessments performed. To make sure these assessments were good, NASA entered into an agreement with the Department of Defense, which has a group (USSTRATCOM) that tracks space objects. With this policy in place, Goddard expanded its service, offering to perform assessments for any unmanned NASA hardware.

The system is heavily automated, performing daily downloads of the latest tracking data and running analyses out to one week for low-Earth orbits and 10 days for geosynchronous ones. Goddard staff perform a more detailed analysis every Monday, Wednesday, and Friday. The software collection uses Matlab, a commercial software package, to run three different types of collision probabilities (2-D, Monte Carlo, and nonlinear). Other commercial software is used to visualize any potential collisions identified by the software.

Potential collisions are flagged for monitoring if there's simply a high probability of conjunction. Typically, the probability goes down after a couple of additional days of tracking, but in rare cases this doesn't happen (and, in a few, the probability went up with further monitoring). When the probability doesn't go down, the software can calculate a maneuver that will reduce the probability of collision to an acceptable level. The solution will take into account other potential hazards as well as mission requirements—some Earth-monitoring satellites can't orbit above a certain altitude and still perform their jobs.

Saving Fermi

This is the system that alerted Fermi controllers that they might experience a close encounter withCosmos 1805, a defunct Russian spy satellite. The expected distance between the two: about 200 meters. For context, the Russian satellite that wiped out an Iridium communication satellite had been expected to pass nearly 600m from it. That and a relative speed of over 40,000 kilometers an hour would be enough to make anyone nervous.

Fermi doesn't actually need to change its orbit in order to function. Eric Stoneking, Fermi's lead engineer for its attitude control systems, said the telescope normally functions in one of two modes. In one, the satellite remains stationary and lets its low-Earth orbit sweep its field of view in order to survey the entire sky. When it is imaging a specific object or event, Fermi rotates in synchrony with its orbit in order to remain pointed at the same location in the sky. Stoneking told Ars that these changes are all done using reaction wheels, the same pointing devices that direct hardware like the Hubble Space Telescope (and Kepler, where their failures are putting the mission at risk).

Since it's mostly just scanning the sky, "Fermi has no orbit maintenance needs," Stoneking said. Despite that, it does have on-board thrusters. NASA requires its satellites to have what is termed "de-orbit capability." Basically, they have to be able to be sent into the atmosphere to burn up safely rather than falling in an uncontrolled manner or staying in orbit and contributing to the junk problem. To deorbit at the end of its mission, Kepler had onboard thrusters.

The problem is that they were only meant to be fired once, after the mission was over. Firing them earlier made people a bit nervous, both because of a history of past problems that occurred while powering up thrusters for the first time—hardware has been lost due to leaks or even explosions—and because of what Stoneking called "the natural inclination to think things over before doing something that can't be undone." Nevertheless, given the 1,400kg threat zooming at it, the Fermi team worked with the Goddard collision assessment group to plan a maneuver that would get it safely out of the way.

Over several days, the risk of a collision did not go away and the team decided to move Fermi. This required stowing the telescope's solar panels and communication antenna, which could have been damaged by the propellant, and using the reaction wheels to point it in the intended direction of travel. When those were done, it only took a one-second firing to move Fermi safely out of the way. With the burn done, the process was reversed in less than an hour and Fermi was back in operation.

Even Fermi's future death was unaffected. "We fired the thrusters for about a second total," Stoneking told Ars. "For deorbit, we'll fire them for well over half an hour."