Re: [find_orb] baffled by longer arcs
Bill Gray Nov 25, 2014
Hi Andy,
I see what you mean.
The problem here is that Find_Orb isn't automatically doing _all_ the
steps one would ideally wish. It's getting a decent first orbit, but
you'll have to do a few things to polish it up.
For example, for (22771) = 1999 CU3, Find_Orb's initial orbit will
include 1403 observations out of 1865, and perturbers from Venus to
Uranus. To get a "final" answer, you'll need to turn on all perturbers,
and hit Full Step so that Find_Orb actually computes an orbit using all
of those perturbers.
Then hit 'Filter Observations'. This will turn on most of the
remaining observations, shut off those with large residuals, and adjust
the orbit to fit the data. This is an iterative process. Sometimes,
after adding in new observations, tossing out those with large
residuals, and computing a new orbit, Find_Orb can look and say:
"Given the new orbit, this observation that was just barely within
the 'maximum residual' tolerance in the Settings dialog is now just
barely outside. And these observations just barely outside have drifted
within tolerance." So it can take a few repeats before things settle
down and Find_Orb reports, "No changes made".
At this point, you should be good. Find_Orb will have an orbit that
includes all reasonable observations, and which includes all the necessary
perturbers. So why doesn't it do that in the first place? Well... you're
right, it should, and I hope to post a version that does all this.
At that point, the "Auto-Solve" button will go away, because the result
Find_Orb shows you will already reflect Find_Orb's best automated guess
at a solution anyway.
I've found (as MPC, Pisa, and Ted Bowell before me have found) that
getting orbits automatically is astonishingly tricky. Some recent improvements
in Find_Orb have put us pretty close to that goal. Lately, I've been
working more on the short-arc cases rather than the long-arc ones; I've
been dealing with issues such as small objects about to fly past us on
short notice, rather than objects that have been observed for years or
decades. But I'm pretty sure I'll be returning to the long-arc case in
months to come.
-- Bill