Jump to content

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

r4m0n

By r4m0n
in KSP1 Mod Releases

 Share

Recommended Posts

sarbian

sarbian

Posted
Posted
For certain values of "updated", yeah, sure. But in reality, Ascent Guidance is borked and MJ has a hard time attaining and maintaining a given attitude.

Can you give more information than just saying it does not works ? Ascent Guidance and Attitude control works fine for me so if you have problem it may be related to a bug with specific ship design or an interaction with a mod.

A screenshot or ship file would help.

Link to comment
Share on other sites
sarbian

sarbian

Posted
Posted (edited)
MechJeb is definitely my preferred co-pilot, but I've recently been using Voneiden's Maneuver Node Improvement for the on-the-fly switching of conics draw mode.

I would love to see this feature implemented into MechJeb so that I could stop using multiple tools :)

It's in MJ since june.

@TimMartland : take the time to read the last 2 3 pages. The same question was asked and answered many time. Get the dev version.

Edited by sarbian
Link to comment
Share on other sites
a__gun

a__gun

Posted
Posted
It's in MJ since june.

No sarbian you misunderstood me - I know there is a maneuver node editor in MechJeb the thing that I think would be a great addition would be the ability to change your conics draw mode from inside the game.

Link to comment
Share on other sites
DerekL1963

DerekL1963

Posted
Posted
Can you give more information than just saying it does not works ? Ascent Guidance and Attitude control works fine for me so if you have problem it may be related to a bug with specific ship design or an interaction with a mod.

Bog standard stock 1.75m parts with no visible vehicle flex, and no mods other than Kerbal Alarm Clock (and the behavior occurred before I installed KAC). This is a vehicle I flew many times in .20 and .21 with no problems.

I seem to have (mostly) solved the ascent wobbles by adding a lot more control authority than I would have in .21, I.E. adding fins, an extra inline stabilizer, and substituting LV-T45's for LV-T35's, though it still oscillates wildly in pitch.... 'hunting' it's target attitude before settling down. (At one point swinging back past vertical.) In .21, with both R4mon's release and the Patch Testbed the oscillations at pitchover steadily decreased in amplitude (and generally only 2-5 oscillations), while with the current dev build they increase in amplitude before decreasing (with as many as twenty plus peaks). Once it finds the target attitude, it flies the remainder of the pitch program as if it were riding on rails.

screenshot9.png

On orbit however, it oscillates wildly with Smart A.S.S. or the maneuver planner (sometimes as much as 180 degrees from it's commanded attitude), and taking sometimes over two minutes to settle down. (As above, in .21 there was generally only a mild overshoot before locking on.)

screenshot11.png

Once it's down to a single stage however... it maneuvers as if on rails. However, reinforcing the staging joints in the core does little to nothing to cure the behavior. Neither does clicking on the Stayputnik and selecting "control from here".

Repeating the above flight with a ".21-alike" configuration (substituting a capsule for the inline stabilizer and science payload, flown many times in .21) exhibited the same excessive oscillations.

Link to comment
Share on other sites
Galane

Galane

Posted
Posted (edited)

Try locking the gimbal on the center engine before liftoff. That's what I learned to do in .21 with MJ. Seems that steady line of thrust not wobbling about helps MJ keep the rocket flying straight.

What I've been having problems with is it not properly finishing orbit maneuvers like periapsis changes and circularizations after the ascent one. It likes to put the throttle down to very low and just keep burning, overshooting the node.

Was testing my 16,000 unit Kethane rover in Mun orbit last night and I had to keep manually stopping it because it would not establish a circular 15KM orbit. I've made some changes to it and will test it again tonight to see if that makes a difference.

Various versions of MJ have done this idling burning, the fix typically seems to be... moar boosters.

Edited by Galane
Link to comment
Share on other sites
a__gun

a__gun

Posted
Posted
Try locking the gimbal on the center engine before liftoff.

Someone PLEASE correct me if I'm wrong but.... Is this not the exact opposite of what the general consensus is? I thought for the best control you wanted to keep vectoring enabled for your central engines and have it disabled for any that are radially mounted.

Link to comment
Share on other sites
sarbian

sarbian

Posted
Posted (edited)
Bog standard stock 1.75m parts with no visible vehicle flex, and no mods other than Kerbal Alarm Clock (and the behavior occurred before I installed KAC). This is a vehicle I flew many times in .20 and .21 with no problems.

Strange, it looks a lot like what I just flew without problem. You run the dev version of MJ or 2.0.9 ? It you run dev did you check the Use Stock SAS in the Attitude Adjustment window ?

Can you post the ship file on pastebin ?

Edited by sarbian
Link to comment
Share on other sites
Starwaster

Starwaster

Posted
Posted
Someone PLEASE correct me if I'm wrong but.... Is this not the exact opposite of what the general consensus is? I thought for the best control you wanted to keep vectoring enabled for your central engines and have it disabled for any that are radially mounted.

That's what I have been doing but after reading up on the Saturn V I intend to try locking just the center. I've only been locking the outers because other people said so. That doesn't mean it's the best method.

I'll say this though, it's hard to roll big craft on fins alone. When gimbals are enabled on outer engines it is easier.

Link to comment
Share on other sites
Galane

Galane

Posted
Posted

Just saying what has worked for me. I have some crazy nightmarish asparagus launchers with one Skipper smack in the middle. They'll wibble all over the place, unless that one Skipper has its gimbal locked.

Look up Pilot Induced Oscillation or PIO. When a flying machine develops a bit of a wobble and the pilot's reaction time is not fast enough and/or overcorrects slightly, attempts to bring the machine under control only make it go more out of control.

The cure is to either keep a firm hand on the controls without trying to do any correcting, or if the vehicle is inherently stable, let go of the controls and allow the vehicle to stop oscillating.

The same seems to apply in this virtual world. A rocket with an oscillation period and/or magnitude that exceeds MechJeb's ability to control with the available control systems on the virtual rocket will constantly overshoot the correction inputs.

Something else that seems to be a control issue with KSP is that aside from the throttle it all appears to be full on or full off with moving aerodynamic surfaces and engine gimbals banging back and forth, lock to lock. Control would be much smoother and gentler on the virtual hardware if there was even one stop between center and full movement. Two would be even better.

What you get with snapping back and forth is parts ripping loose (such as Mainsails under orange tanks) like the tailfin on that Airbus which crashed in New York, October 2011. The flight recorder showed that the pilot was flapping the rudder back and forth, stop to stop, without any pause in the middle as prescribed by the Airbus flight manuals. Would not have happened to a Boeing because Boeings are built to take such abuse by pilots whereas Airbus' design philosophy has long been to design the aircraft to be 'just enough' to do the job - as long as the pilots also adhere to the limits. (Look up the Airbus crashes caused by their initial idea to save fuel by not allowing full throttle takeoffs or full throttle override during landings. "I'm landing here. You don't need full throttle. Oh, hey! Trees!")

So what it looks like to me is that locking the center or central group of engines (or an outer ring of engines) provides a steadying thrust vector that does not wobble about, helping to bring the rocket back onto the desired vector after a control input, while also reducing the need for corrective control inputs because of that always aligned thrust. Sort of like the tailfin and horizontal stabilizer on an airplane provides a constant alignment force while the rudder steers and elevator pitches. (Unless your aircraft is a deliberately unstable fighter jet with a control system and full flying fins that maneuvers by allowing the aircraft to be unstable in the direction the pilot wishes to turn.)

What I can see may be an advantage of locking central engines is the non-varying thrust vector is all aligned directly with the center of mass and having the engines farther from the center do the steering gives them greater leverage and thus a stronger control authority.

Link to comment
Share on other sites
NathanKell

NathanKell

Posted
Posted

Nope, that's the old method of installing addons and is broken. People should really know by now, it's been three revisions since that system was used!

Download this OFFICIAL dev build: http://jenkins.mumech.com/job/MechJeb2/lastSuccessfulBuild/artifact/jenkins-MechJeb2-84/MechJeb2-2.0.9.0-84.zip

and extract it to KSP/GameData/

If you want MechJeb in your pods, download MechJeb and RemoteTech for All, from spaceport.

Link to comment
Share on other sites
DerekL1963

DerekL1963

Posted
Posted (edited)
Strange, it looks a lot like what I just flew without problem. You run the dev version of MJ or 2.0.9 ? It you run dev did you check the Use Stock SAS in the Attitude Adjustment window ?

Can you post the ship file on pastebin ?

The dev version, currently #84. I didn't even know "Use Stock SAS" existed, and it was checked. Unchecked it, and the vehicle flies as expected. What are the consequences of leaving it unchecked?

Edited to add; forgot my manners, thanks for the assist!

Edited by DerekL1963
Link to comment
Share on other sites
drtedastro

drtedastro

Posted
Posted

i am trying to use the .84 dev mod with the new 0.22

Previously i was using Sarbian's version which burnt / used very little rcs fuel. Now, having moved back to the 'original - dev version' of MechJeb, i am burning rcs fuel at an unacceptable rate. I have also tried to go back and use Sarbian version with 0.22, but this is not working for all instances during tests..

is there any 'switch' or way to get this, current, version to quit using rcs fuel so fast?????

thanks in advance for advice / help.

Link to comment
Share on other sites
sarbian

sarbian

Posted
Posted
The dev version, currently #84. I didn't even know "Use Stock SAS" existed, and it was checked. Unchecked it, and the vehicle flies as expected. What are the consequences of leaving it unchecked?

Edited to add; forgot my manners, thanks for the assist!

You're welcome. Strange that it was on by default.

@drtedastro Is it a problem with attitude or translation ? Can you check if "Use Stock SAS" in the Attitude Adjustment window is inactive ?

Link to comment
Share on other sites
Guest
This topic is now closed to further replies.
 Share
Go to topic listing
×
×
  • Create New...