[find_orb] Handling 300 perturbers efficiently

Bill Gray Feb 5, 2013

Hi Jost,

Ah, yes... the wonder days of the Atari ST... brings back some
memories!

Your scheme is probably about what I'd do if I were going to integrate
all 300 (or more) objects against one another. In that case, I'd have
300 * 299 / 2 ways of matching them in pairs.

In this case, though, I'll be integrating one object and looking
to see where it would end up relative to 300 others. I did think about
the "cube" approach (a.k.a. "voxels", the "volume" version of a pixel),
and it _does_ still tempt me. (I've implemented it before for other
problems, and may use voxels in some code to look for linkages between
short orbital arcs.)

I've done a little work on the method I described before, and it's
looking pretty good. It appears to reduce the computational workload
from 300 objects to about 2 objects, on average, in the main belt.
(Move away from there and it drops to zero pretty quickly.) I will
probably not be able to resist the urge to go ahead and plug it into
Find_Orb.

-- Bill

On 02/05/2013 03:02 PM, Jost Jahn wrote:
> About twenty years ago when I have enough time to programming I wrote a
> program for the good old Atari ST to check all virtual (at the sphere)
> and physical (in space) encounters of al known numbered asteroids at
> this time.
>
> My trick was to divide the sphere in many small slightly overlapping
> patches and the space in many small cubes (but not too small) and then
> calculate for one day all asteroids and sorting them into those patches
> in sphere and in space (just the designation). The calculation was very
> rough, about 1/10 precision of those patches. These patches and cubes
> was just 1-byte in a relational file. I count the number of objects up
> and saved the name in a seper ate reational file.
>
> Then I must only make a run through all those patches and each patch
> (byte value>2), which has more than 1 object I calculated much more
> precise. Because I don`t take gravitational perturbances into account
> the result was not good, but nice enough.
>
> Today with computing power I would do the same and when I run throug all
> sphere patches and cubes I would do then (not before) a precise
> calcultion with gravitational perturbations.
>
> I think this reduce the computing time very good. I need only some days
> to run a complete year for all asteroids on the golld old Atari ST
>
> Just my two cents...