The integration errors from IMU state measurements would seem very problematic when you're coming in for a landing, where error tolerances would probably be as small as they could get.

This must mean a state vector update before landing, maybe even several state vector updates at regular intervals? Can someone confirm or dispute?

  • $\begingroup$ "confirm or dispute" exactly what? "Landing" exactly what? From what trajectory? How long before counts as "before"? The answer to "are state vectors updated regularly?" is usually going to be yes, and one of those times may be soon before reentry. Is there any context you can add to this question, it seems to general and open ended. $\endgroup$
    – uhoh
    Commented Apr 6, 2021 at 5:07
  • $\begingroup$ Confirm or dispute that the spacecraft receives one or several state vector updates before/during reentry, from any trajectory, be it from orbit or from reentry post-MECO (like the Falcon 9 stage 1). The spacecraft would be any spacecraft that can land. The integration errors would affect all returning spacecraft coming in to land, since they need to precisely nail down that landing pad/strip/boat, so the problem is general to all of them. Still, I'm not looking for one answer but for as many answers as there might be. I want to have a sense for what is done to mitigate the integration errors. $\endgroup$
    – user39728
    Commented Apr 7, 2021 at 14:40
  • 1
    $\begingroup$ Any spacecraft that can land? That is extremely broad. Soyuz does things very differently than does SpaceX. Vehicles that land on the Moon or Mars (both of which are well-mapped) do things very differently than vehicles that land on the Earth, but also very differently from vehicles that attempt to land on an asteroid. $\endgroup$ Commented Apr 8, 2021 at 19:30
  • $\begingroup$ No worries about differences between spacecraft. I'm open to different ways of doing things. If Soyuz does it one way, it helps to know how they do it, and if the space shuttle did it another way, it helps to know what that was too. I'm interested only in earth landings right now. $\endgroup$
    – user39728
    Commented Apr 8, 2021 at 19:33
  • $\begingroup$ I guess I'm wondering 1) is a 30 m error in 30 s in the ballpark of reasonable for IMU-based state vectors, and 2) how the error is reduced or avoided when landing (either by updating the state vector more often or by using a different navigation technology or some other way). $\endgroup$
    – user39728
    Commented Apr 8, 2021 at 19:36

1 Answer 1


For shuttle, the ground uplinked a state vector about two hours before the deorbit burn.

See the Deorbit Prep Checklist, page 1-13. Action is listed as SV UPLINK.

The vector was updated during entry by data derived from its navigation systems.

For entry, the Orbiter additionally used traditional airplane type sensors: air data probes, TACAN, MSBLS, and radar altimeters. Late in the program, GPS receivers were added and incorporated into the navigation system.


(h/t to David Hammen for the reminder about entry updates)

  • 2
    $\begingroup$ The Shuttle also used navigational clues to update it's state during reentry such as drag, which gave a clue regarding altitude. Lower down, it used TACAN starting at about 700 km from the landing site (and near the end of the Shuttle program, GPS instead of / supplementing TACAN). Modern vehicles use GPS. $\endgroup$ Commented Apr 7, 2021 at 10:21
  • $\begingroup$ @DavidHammen thanks for the reminder! Will update answer. $\endgroup$ Commented Apr 7, 2021 at 12:13
  • $\begingroup$ Ah! So state vector update before reentry and GPS navigation near landing site. But how did they use drag for altitude if drag depends on so many things: velocity, density, drag coefficient, which itself depends on angle of attack, flow regime... Did they have such accurate models of drag coefficient and of density vs altitude (meaning with smaller error than what you'd get from IMU integration)? And wouldn't the calculation assume knowledge of the state vector, the very thing you're trying to update? Curious! $\endgroup$
    – user39728
    Commented Apr 7, 2021 at 14:34
  • $\begingroup$ @OrganicMarble: I'm simulating a state vector estimation algorithm described in an early space shuttle PEG paper, and I'm getting integration errors on the order of 30 m after about 30 s, and this is with the higher nav cycle rate of 25 Hz (as opposed to the 0.5 HZ actually mentioned in the paper for the major cycle). Does this seem unreasonable? 30 m is not much in the big scheme of things, but it's a very big problem when you're doing a vertical landing and you can't afford to shut down the engine 30 m above the landing pad because your altitude read 0 m : D $\endgroup$
    – user39728
    Commented Apr 8, 2021 at 19:11
  • 2
    $\begingroup$ Do you know what a Kalman filter is, @user39728? You haven't asked any questions about them. You are flying blind with regard to guidance, navigation, and control if you don't know what a Kalman filter is. $\endgroup$ Commented Apr 10, 2021 at 19:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.