Project TrajectoryOptimizerMFD open for testing

[Buck Rogers]

I kept it as a way to validate that the ephemeris diagnostic works on others as well. I will either have those two bodies pre-accepted in the release version or actually calculate the residual vel, so that the position and velocity errors fall below the acceptance threshold by themselves.

The whole solution cycle is at least x20 faster than the original version. It really is a blink and you'll miss it. I might add more resolution to the grid as the default setting (instead of 1000x1000 change it to 3000x2000 = 6 times slower). It won't do much to the solution itself (unless you are searching a very long window duration), but it will make it last longer. To be honest, I like it as it is now: Blazing fast.

Blazing fast is great, but one does miss out on calc visuals!

About the "still improving at end" msg; the solution kept improving but you run out of iterations. You can either bump up the population or number of iterations (or both).
If you want to be confident about a multi-body calculation result, you want at least one "nuke" (the solver stalled = it could not get out of the minimum it found, so it "nukes" the population and performs a "smart reseed", keeping the best score history but also trying to get out of the minimum it found).
This is to avoid getting stuck at a local minimum and not finding a better one (that might actually be the global).

So nukes are good!
"still improving at end" was a little hard to decipher. I guess it's hard to make an abreviated text that's easy to interpret, maybe a glossary would be good?

One thing to keep in mind here is that unlike grid searches (body to body), which are deterministic (the result is always the same for a given set of options/constraints), the multi-body calculations are not. They rely on an initial seed and then run a population through a series of calculations that converge on a minimum. You don't actually know if that's the global minimum or if you just hit a local one that the solver simply couldn't get out of. So a good rule of thumb is to have a sufficient population/iterations setting, so that you get at least one "nuke" (preferably two) per run. Also to make at least two runs that produce similar results. Constraining the plan with some variables (v_inf / duration, etc), really works well here and helps the solver converge on the global minimum.

It works incredibly well, it really takes the pain out of doing it "manually".
So one gets feedback on the quality of the solution and when further constraints are advisable.

I know it is annoying but the only other way was to add more buttons to the MFD (new page) which when I tested was even more annoying. My solution is to have "silent" shortcut buttons. When a plan is solved you can see at the bottom : "H = HELP R = Results page". Pressing Shift-R (left or right shift depends on the side of the MFD) will take you directly to the results without going through the TRJ ->BCK route.

No problem, I just thought that as the PLAN button is still there when in plan mode (which makes it obsolete) that it could be swapped out. Not a big deal and with short-cuts noted not a problem.

Kudos on the TransX test. Here is how it should work:
You go to the plan settings and select the "...+ TransX" mode.
When you save a plan it also creates an MFD scenario block that sets the stages in sequence according to your plan. It also sets the v_inf variables for the first leg.
Then when you load the plan, it actually creates a temporary scenario and inserts the TransX MFD block. It loads it (you get the "Importing TransX plan" screenshot and you are now ready to finish the TransX adjustments and fly it (the date is set to 24 hours before the eject date, so you have time to get into orbit, tweak the plan, etc).
This works on Enjo's TransX version and I have only tested in Orbiter 2016.

The .scn re-loading works great, will check IMFD.

The IMFD plan cannot work the same, because it doesn't setup plan as parallel stages. It must be set as a sequence of steps. So instead it will ask you which leg of the journey to load. If it is the first leg, it will set the date to 24 hours before the eject burn, so you have time to prepare. If it is any of the other legs, you continue from exactly where you left off. Note: IMFD doesn't support multi-revolution plans (at least not in the Target Intercept program).

That's why I checked TransX first, IMFD is a little more sensitive, I don't think Target Intercept works with alien systems either?

Both TransX and IMFD cannot support DSM plans.

Sorry, DSM?

About the 229 bodies solar system! Wow that's a nice shake-down of the ephemeris diagnostics! If you can find it, send over a link to the Epimetheus addon. I want to see what's going on there.

Only 7 flagged, and only 3 duds!
Thought you might, I'll PM it to you.

P.S. If you can, try a different reference body other than the SUN. For example a Jovian moons tour or a Phobos-Deimos plan.

Will do

 

[dgatsoulis]

@Buck Rogers

Thanks for the replies and suggestions. Will definitely take into consideration changing the "NUKE" message. I can see how people might think it is a bad thing to happen during a calculation, when it's actually simply an indication that the solver is stuck at a minimum score - which may actually be the global minimum, so no further improvement can be made.

The DSM is a Deep Space Maneuver. You can set it up as a waypoint. It opens up trajectories like Cassini (Earth-Venus-DSM-Venus-Earth-Jupiter-Saturn), or Juno-like trajectories, etc.
You can set the min/max dV for the DSM in the constraints page.
DSMs cannot be setup in TransX or IMFD in advance; you need to actually be "there" (in that part of the flight) to set them up, as they are purely sequential maneuvers.
The MFD will export the plan correctly to TransX (the correct bodies will be according to the plan sequence), but the DSM cannot be set up in advance. Similar thing applies for powered flybies. You have to set those up yourself. The TransX plan export is a convenience that takes away about 50% of the TransX flight plan "chore". It is not a "press button to fly in TransX" feature.