Computers on board spacecraft can determine their positions using a combination of uplinked data from the ground and for those closer to Earth, GPS/GNSS. (See this answer and this question for resources).
The Science Daily article X-ray navigation could open up new frontiers for robotic spacecraft describes NICER's use of timing data collected from millisecond pulsars using its X-ray telescope to generate a solution for it's orbit around Earth, without help. Of course it may use a built-in ephemeris (for the Solar System as well as pulsar rates and epochs) and Earth orbit propagation model (something like SPG4) but I believe it was "blindfolded" and asked to generate it's orbit and real-time position within it using only X-rays.
NICER is currently a payload on the ISS, not technically a spacecraft itself yet, and so it really solves for the orbit and position of the ISS.
Question: I'm wondering if this is the first time that a civilian system in space accurately solved for its own orbit and position using only ephemerides and its own observations in space, without using data transmissions from Earth stations or GPS or other artificially generated navigation signals?
I've mentioned "civilian" a few times because I'd like to exclude discussions of ICBMs etc. as well as inertial navigation. I've added the deep space tag because the utility of X-ray navigation is especially important far from Earth, though it could certainly be used in emergency or unusual situations in cis-lunar space.
You can see several images of NICER (as well as a video) in this and this question.
Excerpts from the article, the full article is certainly worth reading for additional information:
In a technology first, a team of NASA engineers has demonstrated fully autonomous X-ray navigation in space -- a capability that could revolutionize NASA's ability in the future to pilot robotic spacecraft to the far reaches of the solar system and beyond.
The demonstration, which the team carried out with an experiment called Station Explorer for X-ray Timing and Navigation Technology, or SEXTANT, showed that millisecond pulsars could be used to accurately determine the location of an object moving at thousands of miles per hour in space -- similar to how the Global Positioning System, widely known as GPS, provides positioning, navigation, and timing services to users on Earth with its constellation of 24 operating satellites.
"This demonstration is a breakthrough for future deep space exploration," said SEXTANT Project Manager Jason Mitchell, an aerospace technologist at NASA's Goddard Space Flight Center in Greenbelt, Maryland. "As the first to demonstrate X-ray navigation fully autonomously and in real-time in space, we are now leading the way."
This technology provides a new option for deep space navigation that could work in concert with existing spacecraft-based radio and optical systems.
Veteran's Day Demonstration
In the SEXTANT demonstration that occurred over the Veteran's Day holiday in 2017, the SEXTANT team selected four millisecond pulsar targets -- J0218+4232, B1821-24, J0030+0451, and J0437-4715 -- and directed NICER to orient itself so it could detect X-rays within their sweeping beams of light. The millisecond pulsars used by SEXTANT are so stable that their pulse arrival times can be predicted to accuracies of microseconds for years into the future.
During the two-day experiment, the payload generated 78 measurements to get timing data, which the SEXTANT experiment fed into its specially developed onboard algorithms to autonomously stitch together a navigational solution that revealed the location of NICER in its orbit around Earth as a space station payload. The team compared that solution against location data gathered by NICER's onboard GPS receiver.
The goal was to demonstrate that the system could locate NICER within a 10-mile radius as the space station sped around Earth at slightly more than 17,500 mph. Within eight hours of starting the experiment on November 9, the system converged on a location within the targeted range of 10 miles and remained well below that threshold for the rest of the experiment, Mitchell said. In fact, "a good portion" of the data showed positions that were accurate to within three miles.