These are chat archives for esa/pagmo/expsin

26th
May 2015
Chris Andre
@ChrisAndre
May 26 2015 02:01
What is STOUR? I see it a lot in Petropoulos' papers, but I cannot find anything with google
Dario Izzo
@darioizzo
May 26 2015 07:23
STOUR is a propretiary software from JPL which searches "in a grid" for good fly-by sequences.
Roughly works like this: it grids the planet positions. It computes all Lambert arcs between the gridded positions during a tree search. It returns good "close to ballistic" trajectories, candidates to be transformed into low-thrust.
You will see there that, for single revs the pork chops landscape does not change much using expsin w.r.t. Lambert arcs. After we have a good code in place, I would like to reobtain that result as a first check.
Chris Andre
@ChrisAndre
May 26 2015 07:28
The C++ impl. is almost finished
But it does not yet compute all the solutions for 0<N<Nmax
Does it still admit 2N+1 solutions? I'm having trouble conceptualizing it for exponential sinusoids. And I'll try to reason about how to find Nmax soon.
Dario Izzo
@darioizzo
May 26 2015 08:06
Interesting question. I guess you need to look to the $tof, \gamma$ plots to have an answer. I could not find much to say theoretically.
Chris Andre
@ChrisAndre
May 26 2015 10:12
I hadn't seen that report before, thanks!
I'm confused as to how the Lambert solver will fit into the Multi-LTGA problem... Is it assumed that there is an impulse during a swingby?
Dario Izzo
@darioizzo
May 26 2015 10:15
yes ... need to go now, will be back in 30 mins ..
Chris Andre
@ChrisAndre
May 26 2015 11:22
I found this in the report, is this usable for finding Nmax ? >a particular heuristic has been implemented: for a given transfer the number
of allowable revolutions can not be larger than the ratio between the transfer time and the
shortest revolution period between the departure celestial body and the target celestial body
Dario Izzo
@darioizzo
May 26 2015 13:14
It is usable, but its only a heuristic and will probably overestimate Nmax
Could Nmax be computed by computing the max and min tof for each N ? Or this would be too slow?
Chris Andre
@ChrisAndre
May 26 2015 22:35
I've been thinking about how to evaluate the quadrature quickly
The θ˙1 graph is nicely symmetric
My idea is to ignore ϕ and k0 and μ and interpolate between the minima on either side of the origin
and then reconstruct the integral once we have a good polynomial representation of that slice with ϕ, k0, and μ in mind
Chris Andre
@ChrisAndre
May 26 2015 22:41
It might even be possible to cache the interpolating polynomials since the only parameters left to describe it are k1 and k2
I don't know how it will work performance-wise, but we can soon find out!