[PART, 1.0.2] Anatid Robotics / MuMech - MechJeb - Autopilot - Historical thread

[DuoDex]

Sarbian - do you mind if I make a MechJeb derivative that only has the Suicide Burn Countdown? I really like it, but I'm trying to reduce my reliance on MJ for a lot of the things I can do.

[sarbian]

I don't mind, but you know that you can disable whole modules and end up with only the info displays ?

http://wiki.mechjeb.com/index.php?title=Manual/Blacklisting_modules

I should make a cfg for who those who want that ...

[DuoDex]

I don't mind, but you know that you can disable whole modules and end up with only the info displays ?

http://wiki.mechjeb.com/index.php?title=Manual/Blacklisting_modules

I should make a cfg for who those who want that ...

I was not aware of that. I'll be hiding the modules I want to avoid from now on

Thanks!

[djnattyd]

I'm having issues with the ascent AP using the latest Dev build. It isn't performing a gravity turn nor is it applying a roll during climb out. Once at apoasis it isn't turning to face the node (or any node for that matter as SmartA.S.S does nothing.) I can intervene and fly the launch manually with the AP activated and if I point at the node myself, the AP does the burn. The only thing that does work in the Ascent AP is the auto solar panels option. I've updated FAR and the FARExt but the problem with launches is still there.

[DuoDex]

I'm having issues with the ascent AP using the latest Dev build. It isn't performing a gravity turn nor is it applying a roll during climb out. Once at apoasis it isn't turning to face the node (or any node for that matter as SmartA.S.S does nothing.) I can intervene and fly the launch manually with the AP activated and if I point at the node myself, the AP does the burn. The only thing that does work in the Ascent AP is the auto solar panels option. I've updated FAR and the FARExt but the problem with launches is still there.

Gib logs. NAOW.

Can you replicate it?

Does it occur in any other builds?

[smjjames]

And does it happen with multiple ship designs (which would rule out a design issue isolated to one ship)?

I haven't been having those issues, though I'm still getting the overcorrection with the inclination when the ship moves left and right horizontially.

[futrtrubl]

I haven't been having those issues, though I'm still getting the overcorrection with the inclination when the ship moves left and right horizontially.

How do you mean overcorrection? If you have corrective steering MJ will see inclination is too low, it always is, and try to correct it nearer the end.

[djnattyd]

@Duodex logs will have to wait until 7pm GMT as I'm working until then.

@SMJ Only tried one design (but launched 3 times). Craft was (from top to bottom)

Mk16 parachute (realchute nodded)(stage 0)

1 man command pod

1.25m stack separator (stage 0)

The mid-sized 1.25m fuel tank

LV-909 (stage 1)

1.25m decoupler (stage 2)

2 of the long 1.25m tanks

LV-T45 (stage 3)

4 winglets (the ones with the small aileron)

FASA 1.25 Mercury Launch Clamp (stage 3)

Launch profile was set as follows;

Prevent overheat

Terminal velocity

Acceleration (22.9m/s²)

Force Roll (0°) (180°)

Corrective Steering

Gravity turn starts (0.2km)

Gravity Turn ends (69km)

Shape (50°)

Autostaging (stop at stage 1)

First launch went straight up, second and third pitched over to ~85° and stayed there. Once out of the atmosphere it created the node to circularize but didn't turn to face it. Each time this happened I staged until I got to the parachute, turned off the Ascent AP and set SmartASS to 'Retrograde' but it didn't turn to follow the marker.

[futrtrubl]

It's a long shot but do you have KSP installed in Program Files?

[djnattyd]

It's a long shot but do you have KSP installed in Program Files?

It's installed in a Steam folder on a Steam dedicated external HDD

[sarbian]

I will need a screenshot with the attitude adjustment window opened too.

smjjames : I asked those with the overcorrection problem to try a specific build. No one replied => there is no problem.

[smjjames]

I will need a screenshot with the attitude adjustment window opened too.

smjjames : I asked those with the overcorrection problem to try a specific build. No one replied => there is no problem.

I didn't see the post about trying a specific build.

oh i see it, nvm.

Anyhow, I'll check that out.

Okay, heres one with the current dev build, had to turn on limit AoA a few times to get it under control.

Javascript is disabled. View full album

Craft file: http://sta.sh/018bpye1frbc

Parts used are from KWrocketry, Novapunch, SCANsat, DMagic orbital science, MKS/OKS (the Planetary Survey camera comes from this, I think), Karbonite, procedural fairings, Modular Rocket Systems, B9 Aerospace, and Vens stock revamp. Also, just for the record, I'm using KJR.

Edited April 13, 2015 by smjjames

[futrtrubl]

I see you are using FAR. With that you will almost always have to use AoA limiting. It looks like your ship, at the speed and fuel loading when it occurs, has a positive Cm at small AoA which goes negative at large AoA meaning you don't loose complete control but find it difficult to maintain stable flight.

Also I don't recommend using corrective steering, especially with FAR. I have done tests and there is never a situation it reduces dV needed for launch and it tends to try point you in odd directions later in the ascent. The only thing useful it could do is correct inclination during launch since MJ always ends up with an inclination less than it was asked for.

Edited April 13, 2015 by futrtrubl

[smjjames]

I see you are using FAR. With that you will almost always have to use AoA limiting. It looks like your ship, at the speed and fuel loading when it occurs, has a positive Cm at small AoA which goes negative at large AoA meaning you don't loose complete control but find it difficult to maintain stable flight.

True, but that behavior is not there in build 411, yes it bucks up and down a bit, but doesn't have that massive yawing (I think thats the right term?) left and right.

Also should test it without the corrective steering for the current build. I'll upload the album for the build 411 test soon.

Usually the only reason why I put AoA on is to keep it from flipping backwards or turning forwards too fast, not from doing the left and right behavior.

Javascript is disabled. View full album

Edited April 13, 2015 by smjjames

[sarbian]

Ok, so 411 works. That means Xytovl feed forward patch is the most likely culprit. I had an other problem with it while testing some stuff with the landing AP so I guess I will roll it back for now. Expect a build later today.

[smjjames]

I did some further testing and things went pear shaped a few times. I think there MAY be a design or balance problem with the craft, or maybe it really is MechJeb, I'm not totally sure. I'll upload the album of my further testing soon.

LOTS of images in this one.

Javascript is disabled. View full album

Edited April 13, 2015 by smjjames

[futrtrubl]

I did some further testing and things went pear shaped a few times. I think there MAY be a design or balance problem with the craft, or maybe it really is MechJeb, I'm not totally sure. I'll upload the album of my further testing soon.

LOTS of images in this one.

http://imgur.com/a/O1Csb

Don't forget you can click "show navball ascent path guidance" to see where your ship wants to be pointing.

[sarbian]

And you can activate the "Requested Attitude" arrow (unselect the "arrow origin at the CoM" so the arrow starts at the command pod)

[smjjames]

Don't forget you can click "show navball ascent path guidance" to see where your ship wants to be pointing.

I did a couple of times during the tests to see what it was doing.

And you can activate the "Requested Attitude" arrow (unselect the "arrow origin at the CoM" so the arrow starts at the command pod)

Oh I should have unselected 'arrow origin at the CoM' for better analysis? Didn't realize that.

[sarbian]

It just change where the arrow is shown. At CoM make it easier to understand CoM related stuff and Not at CoM make it easier to see it since it's not buried in your ship (or you use the visible though object like you did).

The attitude (orientation) stuff use the active comment pod as reference, so it make a bit more sense to use that too.

[smjjames]

Well, hopefully that analysis helps you find out what MechJeb is doing wrong.

When doing build 411, I nearly hit the limit AoA button since it pitched up a bit early on, but it held.

[xytovl]

Ok, so 411 works. That means Xytovl feed forward patch is the most likely culprit. I had an other problem with it while testing some stuff with the landing AP so I guess I will roll it back for now. Expect a build later today.

With feed forward, the controller will be much more reactive to target changes. I don't know what the corrective steering is supposed to do exactly, but it has certainly been tuned with the "normal" attitude controller in mind.

If anyone knows the control it should implement, we can fix it, otherwise we may drop the feature.

[sarbian]

Xytovl : If you want to see the problem I had put the top stage of a Kerbal X on orbit around the Mun and use the landing AP to land on a specific target. At various point the lander kept chassing a vector and the lander lander on its side. I worked without the change

My current take is that the "no feed forward" code works fine in Vacum. In atmo we still have some problem with feed forward in some case, but I did not isolate a clean pattern.

Not sure I ll have the time for a build today...

Edit : I just switched the file to its old state and build. It should go back to the old behaviour for now.

Edited April 13, 2015 by sarbian

[smjjames]

I was actually having similar difficulty with large ships that I was attempting to land, where it would sort of land sideways or otherwise crash into the ground.

Still though, the ships that I was attempting to land had some bad engine and RCS balance issues, so that (coupled with the weird and oftentimes flailly RCS behavior) likely contributed to the problem. So I can't completely rule out the problem not 100% MechJebs fault. The current iteration of that ship is much better balanced, so I'll see how that goes.

Maybe I'll land that and an earlier iteration using build 411 and see how they perform.

Or rather I'll use 429 since there's some improved behavior between 411 and now.

Edit: Does the orbit height matter for the Kerbal X example?

I tried a few times with the Kerbal X example and didn't end up landing on it's side, though it did chase some vectors while in the final landing sequence. I probably also don't know what to look for as far as what it should be or shouldn't be doing.

Edited April 14, 2015 by smjjames

[smjjames]

Has anybody had problems with the orbit info, Delta V stats, surface info, and vessel info windows not work (as in not appearing when you toggle them in the sidebar, though they work through blizzys toolbar) after you dock and then later undock a craft from a station (or whatever you're docking with)? I've run into this multiple times already.

I'm not seeing any accompanying exceptions in the output log or exception detector.

Edit: To clarify somewhat, those are working for the sidebar when I'm focused on the craft that I docked with, but not the origional craft that I undocked.

Edited April 15, 2015 by smjjames