In article ,
Danny Dot wrote:
Does anyone know how we navigate our deep space probes? I know we use radar
and GPS in LEO. What do we use for deep space? My guess is we use Doppler
shift a lot -- but that is just a guess.

The dominant method is radio tracking -- radar, but with the assistance of
the probe to repeat the signal back. (Normal radar is useless for such
small objects at such distances.) Round-trip time gives range, and Doppler
shift gives range rate (velocity along the probe-Earth line); these can
both be measured extremely accurately. More subtly, the *change* in
Doppler shift as the Earth rotates can be used to determine direction,
although not as accurately. These measurements are combined using a
sophisticated estimating process that looks at how the measurements
*should* be changing for a given trajectory, and solves simultaneously
for the most likely probe position/velocity and the most likely values of
other parameters (masses of planets, errors in measurements, etc.).

Some other techniques sometimes get added. If multiple ground antennas
are available *and* the probe's signal is a strong one, interferometry can
be used to measure direction *extremely* accurately. Maneuvers can be
measured by onboard accelerometers for comparison with tracking data. And
if the probe has a suitable camera, images showing planet(s) against a
star background can be used to establish position; this "optical
navigation" is especially useful when approaching a planet, because it
gives an independent check on position relative to the planet, which is
what you mostly care about.

My understanding is the navigation system of the probe that had the unit
conversion error was "telling" us that there was a problem, but we didn't
have enough confidence in the navigation system to realize there was a
catastrophic error in the trajectory.

Kind of, sort of. That was Mars Climate Orbiter. The units error messed
up the estimation of a growing trajectory error caused by small thruster
firings. The probe itself was not doing its own navigating. It *was*
computing better thruster-firing estimates than the ones produced on the
ground, but the navigation team wasn't aware of this and wasn't using the
results.

The MCO radio-navigation data wasn't behaving *quite* as it should have,
but there were other problems with using the data in the beginning, the
error happened to be in a direction that routine tracking wasn't very
sensitive to, and the small anomalies weren't conspicuous enough to alert
the navigators. (People *did* notice that something wasn't quite right,
but heavy workloads prevented prompt investigation of problems that didn't
seem serious.) VLBI wasn't used because the antenna time needed for it is
not simple to arrange, and MCO was seen as a routine mission that didn't
need special measures. And although MCO had a small camera, it wasn't
really very suitable for optical navigation and so none was attempted.

At the last minute, as MCO approached Mars and Mars's gravity began to
affect it, the discrepancies started to grow: Mars wasn't affecting the
trajectory quite as expected. The navigators started making corrections,
but they were busy and didn't have time to investigate properly, and the
severity of the underlying error -- and thus the size of the correction
that was really needed -- wasn't obvious until it was too late.

The real problem wasn't the units error, but the overworked team, the
persistent failure to investigate small but unexplained discrepancies, and
complacency based on the belief that entering Mars orbit was routine.
--
spsystems.net is temporarily off the air; | Henry Spencer
mail to henry at zoo.utoronto.ca instead. |