> > [..]
> > The road map we've been implicitly following is this:
> > . LV2 demonstrates telemetry and orienteering
> > . LV2+ demonstrates trajectory following
> > . LV3 design phase low cost amateur launch vehicle
> > . Motor shopping
> > . LV3 orbital operations
> any plan for multi-stage testing?
Presumably LV3 is a 2 or 3 stage vehicle. It's still all very
preliminary but i would assume flight testing the upper stage
1st. Possibly flight testing the lower stage separately, Then
doing a multi-stage orbital (woo hoo) mission.
> > [..]
> > Crunched versions of LV2 have demonstrated telemetry and sensor
> > package. (Despite the transverse noise i'm not worried about the basic
> > IMU design.)
> > IMNHO we know how to solve the orienteering problem. If we had a
> > couple full time graduate slaves we could bang out a solution in a
> > couple months.
> how  will the orienteering problem be solved?
Well, this seems like a fair question ;)
Nowadays it's almost been reduced to a textbook exercise, but not
For an opening paragraph i get to pick buzz words, i think i'll pick
"sensor fusion" and "complementary filter".
We want the control system to do several things
. Follow a prescribed trajectory with multi-meter accuracy
. Orient the vehicle with ~degree accuracy
. Damp structural instabilities
. Respond gracefully to un-programmed events
. Respond quickly if appropriate
. Be reliable and fault tolerant
In addition we want the telemetry to report vehicle orientation with
~meter/~degree accuracy at sub-second rates.
I think all these points are defensible, for now i will assume them.
Formerly solving a tracking problem like this would involve something
like the Goldstone radio antenna or similar infrastructure. These days
we have satnav like the GPS. For meter-level accuracy the GPS is
sufficient, but has several short comings
. Low update rates
. High drop-out rate
. Does not provide attitude information
Actually, all of these problems can be addressed. It's even
possible to do a GPS-only sensor package, but it's probably not
wise. Rather the approach we've been pursuing is to get the best GPS
unit we can, cheaply. That means update rates of ~10Hz, probably no
multi-second dropouts, and, at least at 1st, no GPS attitude info.
Therefore the GPS is insufficient on several counts
. Too slow
. Not reliable
. No attitude information
Maybe i'm being too pedantic here, because these are clearly the
reasons why we build the IMU. The IMU has properties that complement
the GPS. Specifically
. Fast (100's Hz)
. Reliable (No outside signals)
. Measures complete inertial state
The IMU also has the significant problem of drift. After a certain
number of seconds position estimates derived from IMU data are
inaccurate due to integrated measurement error.
The idea then is to use the GPS with meter position accuracy and
sub-second un-reliability coupled to the IMU with multi-second
accuracy and high, fast precision. Together the two systems are
meter/degree accurate, kHz fast, and somewhat fault tolerant.
To do the actual sensor coupling we are currently pushing two
different approaches. The more conventional approach is the extended
Kalman filter (EKF) which basically does a least-squares estimate on a
linearized state-space model. The other tack is a Bayesian particle
filter (BPF). This is a Monte Carlo method operating in real time on
a system model that does not necessarily need to be linearized.
There is a trade off between the number of sensor inputs, reliability,
and computational burden. So far we have generally chosen to pursue
sensors which are cheap and orthogonal. Here are our favorites in
. accelerometer / rate gyro (aka IMU)
. telemetry Doppler
. other radio nav
. active radar
. star sensor
. camera vision
I'm probably forgetting some, we really should have a list
somewhere. Anyone care to add?
Currently we're planning to use everything from magnetometer upward on
psas-software mailing list