NavList:

   

Reply

James R. Van Zandt wrote:
> 
> The main technical obstacle is the calculation of all the partial
> derivatives.


For my satellite orbital element differential adjustment program, I
obtained the partial derivatives numerically. The program prompted for a
∆ to apply to the element being adjusted, say inclination. It then
displayed the corresponding change in the satellite's predicted position
(in arc minutes) at each observation, and asked if that was acceptable.
I aimed to make it about the same order of magnitude as the residuals.

The program could do a simultaneous least squares adjustment to any
desired combination of orbital elements while holding the others
constant. Observations were independently weighted according to the
estimated arc minutes of accuracy supplied by the user. At the end of
each adjustment run, residuals for the observation set were displayed,
as well as an overall figure of merit. It was up to the user to decide
if another iteration was necessary.

The number of observations was limited only by available memory. In
practice, this was no limitation at all. An artificial satellite
orbiting Earth has very "bumpy road" compared to a planet orbiting the
Sun. It's affected significantly by atmospheric drag and complex
irregularities in the gravity field. The orbital models (I used NORAD's
SGP4 model) allow for these effects in the short term but not over a
period spanning several years. To generate accurate elements, you must
use fresh observations.

My least squares formula came right out of a textbook. Later I read that
this elementary formula is sensitive to error, and not recommended for
demanding applications. However, in my program it converged rapidly to a
stable solution. I never felt the need for anything more sophisticated.

I used this program to maintain orbital elements for some U.S.
reconnaisance satellites. The Pentagon won't release the elements for
those birds, but a number of hobbyists track them visually. My elements 
for the Lacrosse radar satellites were so accurate, you could set 
binoculars on a predicted spot against the star background and watch the 
bird come sailing through on course and within a few seconds of the 
predicted time.

All this was done in the C language on an 8-bit computer with 64 k of
RAM and a 2.5 MHz processor. It was a long time ago -- I no longer have
a machine able to read my old floppies. But when I read Kaplan's paper I
recognized many parallels to what I used to do.

-- 
I block messages that contain attachments or HTML.




--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to NavList@fer3.com
To , send email to NavList-@fer3.com
-~----------~----~----~----~------~----~------~--~---

   

Reply