NASA OS Software: The General Mission Analysis Tool (GMAT)

[ElChristou]

GMAT only work for non atmospheric trajectories, right?
No way to generate the reentry of the capsule or the SLA panels or any others abandoned pieces? What about the Moon? possible to show the crash trajectory of the ascent stage of the LEM (after comeback of astronauts in the CSM)?

[Fenerit]

Then the correctness of lemniscata shape of the path was perfectly realized with an old '69 computer? Well done, we can say about.

[Fenerit]

Sorry, I'm wrong. With an '69 engine attached to a '69 computer.

[t00fri]

...
To repeat: accuracy for the planets, accuracy for the moon
(which is a special issue!) , easyness in handling AND
speed considerations. And more recently, there is the fact
that SPICE and GMAT do NOT use VSOP87.

...
Bye Fridger

Do you remember roughly how different the VSOP87 and
VSOP2000 were for moon position, and how much they
differ from the DE40x files?

djcinsb,

my quantitative comparisons for Celestia lie ~ 5 years
back, and hence quite a few details of our findings have
been forgotten. Yet some remained

1) We never compared the VSOP87/2000 predictions with
each other or against DE40x.

Before turning to the moon problematics below, let me
summarize for you some typical orbit tests I did with
Celestia:

I mainly checked the overall orbit precisions against a host
of so-called "mutual event" timings. The most precisely
tabulated such events are a host of detailed eclipse
timings by famous Jean Meeus and his team among the
Galilean moons of Jupiter. These striking eclipses among
the "tiny" moons were reproduced by Celestia within
~1-3 seconds of time (!), after accounting for the
light-travel delays, of course.

Here are a few of those striking comparisons I made in
2002:

1) The /annular/ eclipse shadow of Europa on Io (2e1A)
on Dec. 27 2002



2) The /total/ occultation of Europa by Callisto (4o2T) on
Dec. 15 2002



This was also a striking confirmation after I had located a
very nasty hidden bug in our VSOP87 code ...

Then I went on with testing many analogous events among
the Saturnian moons that were also accurately available in
form of precisely timed earthbound photographs. Here the crucial tests
refer mainly to high-precision timings of ring-plane
crossings by the moons.

... analogously with moon-events for Neptun and Uranus.

Finally I did many striking tests with the Pluto-Charon
system: During the period from 1985 through 1990, Pluto
and its satellite Charon underwent a series of transits,
eclipses, and occultations, which again are collectively
called "mutual events."

Here is a graphical display of many respective
high-precision observational data



The respective Celestia timings were again excellent!



Or e.g. this detailed comparison



2) By inspecting Celestia's solarsys.ssc data file, you will
note that VSOP87 was applied to the major planets, while
the Moon still is a "primitive" elliptical orbit. So here
would certainly be much room for improvements along
VSOP2000, for example.

As is well known, the VSOPxx approach is a
semi-analytical approach in terms of expansions in
terms of hundreds of Chebyshev (orthogonal)
polynomials. The coefficients were best-fit to the JPL
DE200 and successive JPL DE450 numerical-integration
models. So that's the general relation between the
VSOPxx and JPL DExxx approaches. Clearly the price for
the numerical integrations is high in the case of real-time simulations
like Celestia, since for reasons of speed HUGE data files
have to be carried along! They presently range between
50MB and 200MB as you know well. This aspect probably
does not touch GMAT very much, since I understand that
you are mainly interested to generate very precise
optimized orbits, yet NOT in real-time.

For the Moon, ELP2000-82 and its successor ELP2000-82B
are semi-analytical models, again with fine-tuned
parameters versus JPL DE200.

So here we should definitely invest some work by
replacing the present moon code by the ELP2000-82B or
even VSOP2000 semi-analytical expressions.

I hope this gave you some idea about what was done as to
the various orbit approaches and their precision tests in
Celestia.

Bye Fridger

[djcinsb]

Fridger -

Thanks for taking the time to post such a detailed reply. It looks like you a lot of hard -- and excellent -- work here!

I am concerned about the Earth-Moon system, but that is mitigated quite a bit by the option Selden mentioned to use the JPL ephemerides in Celestia 1.5 -- once I figure out how to get that version (a beta copy for now, I assume).

When I was playing around with the Apollo 11 transfer on Friday, I noticed a difference in the plot of a GMAT calculated Moon-centered trajectory from an Earth-centered trajectory, on the order of 100 km (judging by appearances in Celestia). Both trajectories come from the same data in GMAT, and they do transform back and forth correctly inside of GMAT. Both are built using DE405 data. The difference as seen in Celestia is probably coming from the difference between using the DE ephemeris and the elliptical orbit in Celestia 1.4.1. It is a large enough difference to make a trajectory targeted in GMAT to an apoapsis location of 183 km appear in Celestia to be much closer to the Moon (it looked like the spacecraft was coming in around 100 km away at periapse), and to make the apoapsis point show up slightly earlier for the Apollo 11 encounter July 19, 1969. Note that I don't think of this as a major issue -- GMAT and Celestia are solving different problems, so I wouldn't expect us to agree on every item! And the potential update to set the Moon state a bit more accurately using the VSOP2000 sounds promising. As does the option for using JPL ephemeris files for special case users like GMAT folks.

I do understand the performance -- and download size! -- issues you raise about using the DE files. While we can boast that "GMAT is thousands of times faster than real time," that means something very different for flight dynamics problem solving (after all, we do model a precise trajectory from Earth to Mars in under a second, while flight time is around 7 months!) than it does for the types of visualization and precision body location updates needed in Celestia, where you have to update the graphics many times a second!

Thanks again for the information filled response. I hope that we'll be able to continue to pursue things together -- if we can ever figure out what that means!

- Darrel

[selden]

Darrel,

Celestia does use a CustomOrbit for the Moon. The EllipticalOrbit definition is included in solarsys.ssc for illustrative purposes, just as it is for the planets using VSOP87 theories. I don't know the source of the CustomOrbit's values, though.

The installer for Celestia v1.5.0pre3 for Windows is available at
http://www.celestiaproject.net/~claurel/celest ... .0pre3.exe

A copy of the Celestia executable (not the installer) built by Vincent for Windows from the current CVS source code is available at
http://vincent.gian.club.fr/celestia/ce ... .5_CVS.exe
(but it does not include spice support)

Steve Binder has made available a build of pre3 for the Mac. See http://www.celestiaproject.net/forum/viewtopic.php?t=11096

Alternatively, you could download the source code from the cvs archive on SourceForge and build it yourself.

[t00fri]

Darrel,

Celestia does use a CustomOrbit for the Moon. The EllipticalOrbit definition is included in solarsys.ssc for illustrative purposes, just as it is for the planets using VSOP87 theories. I don't know the source of the CustomOrbit's values, though.

Oh yes, now I remember about the Moon.

I just looked at the code in customorbit.cpp.

The present method to compute the lunar position is pretty complex. Chris must have tacitly modified the Moon orbit completely, as often WITHOUT any reference to where those many terms have been taken from!

Here are his CVS comments from about a year ago, when the overhaul has tacitly happened:

Fixed and improved interface to JPL ephemerides:
- The orbit of the moon was completely broken; fixed this
- The orbit called 'earth-jpl' was actually using the position of the Earth-Moon barycenter; changed it to use the position of Earth and added jpl-emb for the Earth-Moon Barycenter
- Added heliocentric and barycentric orbits for planets
- Added geocentric and barycentric orbit for Moon

The last mod of that file was Nov 6 2006.

In 2002 when many tests were done it was explicitly said:

Use an elliptical orbit for the Moon outside a range of 2000 years from present

From various other comments one may guess that Chris inserted the VSOP87b solution also for the moon (using the equinox of J2000 instead of the equinox of date).

Yet I never did a lot of tests there since I felt that the moon orbit should anyway be improved by adopting another series expansion specializing on the Moon ephemeris.

Bye Fridger

[steven hughes]

Steve,

For those who aren't on the best of terms with celestial mechanics , programs like this can be intimidating. What's the chance of the documentation including a "worked example"? e.g. takeoff from Earth through landing on Mars?

It's been awhile since I posted to this thread. I thought I would get to putting together some tutorial info sooner, but I've had a really hectic month. Here is an example of where we're going with the tutorials. The link below illustrates the first of many others to come...

http://gmat.wiki.sourceforge.net/Creati ... Spacecraft

[selden]

Steven,

That looks like a great start.

Thanks!

[steven hughes]

I've completed the first tutorial I mentioned in a previous post. Any feedback is welcome .... clarity, missing info, needed definitions, etc.

http://gmat.wiki.sourceforge.net/Creati ... Spacecraft

I'm going to use the Apollo mission as the basis for a Lunar Transfer tutorial soon... although that may be jumping a few steps in complexity, we'll see.

Darrel is working on a Celestia subscriber in GMAT, so eventually, you will be able to output XYZ files from GMAT, so Celestia can plot them. Not sure when that will be ready though.

[mjoubert]

Great job. Everything is well explained.