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

[johnsonwax]

Interesting info. Looks like I need to rethink some of these top heavy rockets huh?

What has helped me considerably is two things:

1) Ditch the conventional design. I design with my orbital stage in the center at the bottom. Often a nuke drive or low thrust high vacuum efficiency motor - 3.75m fuselage. Surrounding that are 3-6 (depending on size and design of payload, but usually 4) large boosters. These are usually either 3.75m or 5m single or dual stage with a TWR of 1.5-2.0 and a combined dv of about 4500. Atop the uppermost booster stage is a truss that extends vertically to the height of the payload. This arrangement reduces the height of the rocket considerably by lowering the orbital stage as low down within the boosters as possible (I cut the height of my rockets by 30% or so), and provides additional means to strut the top end of the payload to the main booster stages (much less wobbling). If designed properly, the upper booster stage will run out of fuel at the orbital insertion with a periapsis within the atmosphere. The boosters then break away (sepratrons) and can reenter and be reclaimed, but it keeps the entire craft rigid and stable until its out of the atmosphere. The orbital stage can then do the last 100-200 dv to circularize and go from there. You can use the orbital motor as part of the lift if that helps (usually it does) and have the boosters feed fuel to that tank so it arrives topped off. I can reliably launch 140t with this setup including 50m long station parts.

2) On those boosters add a small rocket to each pointing radially outward at about 45 degrees. I use 20-75 power rockets (like the little orange guys) for this. They use very little fuel but by moving some of the thrust inward along the center of lift axis, it helps with spin and attitude control - just enough to help the gimbals. Make sure they are pointed dead center using the snap adjustment, though. I also add some movable canards on the outer surfaces. They're light and help with attitude control early in the lift but don't help much up around 30KM which is when a lot of large rockets tend to go off - as the terminal velocity really increases.

[krashkart]

Is there some special thing that I have to do to get MJ2 to work with 0.21.1? The part won't attach to anything and it breaks the VAB until I start a new craft.

Edit: It's the only addon that I have installed at this point.

Edit2: Resolved. In case anyone else is having the same problem, try using sarbian's DLL.

Edited August 21, 2013 by krashkart

[synik4l]

What has helped me considerably is two things:

1) Ditch the conventional design. I design with my orbital stage in the center at the bottom. Often a nuke drive or low thrust high vacuum efficiency motor - 3.75m fuselage. Surrounding that are 3-6 (depending on size and design of payload, but usually 4) large boosters. These are usually either 3.75m or 5m single or dual stage with a TWR of 1.5-2.0 and a combined dv of about 4500. Atop the uppermost booster stage is a truss that extends vertically to the height of the payload. This arrangement reduces the height of the rocket considerably by lowering the orbital stage as low down within the boosters as possible (I cut the height of my rockets by 30% or so), and provides additional means to strut the top end of the payload to the main booster stages (much less wobbling). If designed properly, the upper booster stage will run out of fuel at the orbital insertion with a periapsis within the atmosphere. The boosters then break away (sepratrons) and can reenter and be reclaimed, but it keeps the entire craft rigid and stable until its out of the atmosphere. The orbital stage can then do the last 100-200 dv to circularize and go from there. You can use the orbital motor as part of the lift if that helps (usually it does) and have the boosters feed fuel to that tank so it arrives topped off. I can reliably launch 140t with this setup including 50m long station parts.

2) On those boosters add a small rocket to each pointing radially outward at about 45 degrees. I use 20-75 power rockets (like the little orange guys) for this. They use very little fuel but by moving some of the thrust inward along the center of lift axis, it helps with spin and attitude control - just enough to help the gimbals. Make sure they are pointed dead center using the snap adjustment, though. I also add some movable canards on the outer surfaces. They're light and help with attitude control early in the lift but don't help much up around 30KM which is when a lot of large rockets tend to go off - as the terminal velocity really increases.

Dang. Very good info. Thanks man. Could you by chance send me a pic or two of what you mean. I think im visualizing it correctly. but i may not be.

[Tassyr]

Hey, has anyone figured how to add this to the stock pods again? I'm trying, but whenever I do they kinda break. The pod just stays -exactly- where it is at launch, gravity and physics be damned...

[GavinZac]

Hey, has anyone figured how to add this to the stock pods again? I'm trying, but whenever I do they kinda break. The pod just stays -exactly- where it is at launch, gravity and physics be damned...

http://kerbalspaceprogram.com/mechjeb-and-remotetech-for-all/

[magnemoe]

Interesting info. Looks like I need to rethink some of these top heavy rockets huh?

edit: and when I say top heavy...I just mean things hanging out from the top...causing it to pull more from the top. making it alil harder to keep control

My standard design is pretty much boosters around the central payload for heavy loads.

Rocket is more stable, its also far more structural stable, the boosters actualy help keep the payload together.

As the heavy loads tend to be interplanetary crafts with nuclear engines I can let them help me to orbit, it also let me use the boosters as drop tanks for circulating and start of interplanetary burn.

[sarbian]

I am back. You have me for 3 weeks before I get a 5 weeks break

Still no reply from R4m0n so I'll proceed with a new thread later today.

[GavinZac]

I would say MJ works great just landings for me are way off course but I guess they will fix this and yes I am doing some minor course correction manually

I would suggest not doing minor course corrections, as MechJeb will then just want to correct them back to the original plan. MJ2 is very accurate for me, in fact sometimes too accurate, I forget to change the target by a few metres and end up landing on top of a previous craft.

Are you deploying parachutes? That's one thing MechJeb doesn't take into account.

[FredMSloniker]

I thought I read somewhere that MechJeb had recently added the ability to align docking ports when docking, but if it's an option I can't find it. Is that something planned, or did I remember wrong?

EDIT: by 'align', I mean rotationally.

[Simplicity]

I am back. You have me for 3 weeks before I get a 5 weeks break

Still no reply from R4m0n so I'll proceed with a new thread later today.

For the lazy ones could you possibly add the new thread link you going to make in your signature for us please

[sarbian]

For the lazy ones could you possibly add the new thread link you going to make in your signature for us please

I will, but it will be later this week since RL is busier than expected ...

@FredMSloniker : DLL in my sig does it

[johnsonwax]

Dang. Very good info. Thanks man. Could you by chance send me a pic or two of what you mean. I think im visualizing it correctly. but i may not be.

I'll put up some pics of the launcher in a few days. Check my blog - I'm doing some tutorials now and the ship will show up shortly.

[triffid_hunter]

I didn't see this suggested anywhere, but if it has, maybe point me to a discussion. I'm wondering if a feature could be added where a launch could be timed to the position of another vehicle already in orbit, either by a direct time measurement or by the moment the (selected) orbiting vehicle passes a reference coordinate on the surface.

What I'm trying to do is set up a comm relay network where the relay sats are equally spaced in their orbit. I think I know the basics of doing it (manually), but it's difficult to "push the go button" at the precise moment, since I can only watch one spacecraft at a time. The time it takes to switch views from the orbiting sat back to the ground, engage the ascent computer, then make it go, moves the orbiting sat farther ahead than what I want.

MechJeb can already do this. Target the craft you want to rendezvous with, then at the bottom of the ascent autopilot there's "launch to rendezvous with target".

This relies on the LPA being correct- the easiest way to get it right is to do a launch with the autopilot, then revert after the circularisation burn is complete and the autopilot deactivates itself.

LPA stands for "launch phase angle", and is simply the angle between KSC and your craft when the circularisation burn is complete. This tells MechJeb how far in front of your target to launch.

If you want to place a craft at a particular angle vs another craft, simply add that angle to the LPA before hitting launch to rendezvous!

If the orbital periods aren't exactly the same, they'll slowly drift. One of the instrumentation dialogs shows you your orbital period.

The one aspect I've had to take over from Mechjeb on more times than I'd like to count recently is docking. It just can't do it right. Either it swings about wildly, never correcting until it has swung almost 90 degrees off course, or lining up perfectly but never using the translation controls.

I don't even use the docking autopilot these days. Instead, I set Smart A.S.S. to PAR- then use engine burns or translation to make the target prograde marker push the pink target sigil towards the center of the navball. This is waay faster and massively more efficient on RCS than MJ's docking autopilot.

I thought I read somewhere that MechJeb had recently added the ability to align docking ports when docking, but if it's an option I can't find it. Is that something planned, or did I remember wrong?

EDIT: by 'align', I mean rotationally.

If rotational alignment is important to you, use four regular ports on a quad adapter like this. Make sure to roll into proper alignment prior to docking though

[johnsonwax]

If rotational alignment is important to you, use four regular ports on a quad adapter like this. Make sure to roll into proper alignment prior to docking though

THSS and KOSMOS also have positive lock docks. You can only dock at fixed angles. Makes docking fussier in the last meter, but the results are very satisfying.

[ninjaneer]

Small notes on 2.0.7:

-- Docking autopilot has worked so far for me.

-- Don't know if this is planned, but the status line in the ascent autopilot (the line of text that says what's going on -- gravity turn, coast to apoapsis, etc.) isn't displaying, except very very briefly after engaging the ascent AP.

I don't know where this thread has been since I've been busy playing, but my experience has been:

* Ascent and rendezvous autopilots are awesome and extremely optimistic, and conserve a lot of fuel

* Docking auto is extremely pessimistic and will burn through 1000 units of mono in no time just to "be perfect" on z 100m out. The game is much more lenient on that, and despite the waste it still won't align xyz with target angles on dock, resulting in many docks an asterisk formation. Auto-docking should also use SRS/burn if it can and take a "good enough" approach since it seems to feel that way about docking in general given the failed angle approach. Especially when the port is lateral to the target/craft rendezvous approach.

[sojourner]

1000 units of monoprop??? What are you trying to maneuver - an aircraft carrier?

I typically use the rendezvous autopilot to take my ship into 50 meters, then use Smart ASS on both ships to point the docking ports at each other (this saves tons of monoprop since Smart ASS will use reaction wheels), then I can either finish the docking manually or let docking autopilot do it for much less than 100 monoprop.

In fact, I just docked two orbital taxis using this method and the docking autopilot used only 1 unit of monoprop. total.

Edited August 22, 2013 by sojourner

[plasman]

I don't know where this thread has been since I've been busy playing, but my experience has been:

* Ascent and rendezvous autopilots are awesome and extremely optimistic, and conserve a lot of fuel

* Docking auto is extremely pessimistic and will burn through 1000 units of mono in no time just to "be perfect" on z 100m out. The game is much more lenient on that, and despite the waste it still won't align xyz with target angles on dock, resulting in many docks an asterisk formation. Auto-docking should also use SRS/burn if it can and take a "good enough" approach since it seems to feel that way about docking in general given the failed angle approach. Especially when the port is lateral to the target/craft rendezvous approach.

Mechjeb seems to have great difficulties to pinpoint the docking port. So yes you have to previously align both vessel's docking ports in order to avoid the docking system wasting tons of monoprop. My method for docking is very similar to sojourner's.

-Use the rendez-vous autopilot to approach within 40-50 m of the target

-Align both vessels

- Engage Mechjeb.

- Disengage Mechjeb within 10 m of the docking port. Use the RCS to realign. Generaly Mechjeb does a good job at approaching near the docking port, but his aiming is just not good. So you have to realign by translating your vessel (RCS' keys J-L and K-I). Once properly aligned switch on the SAS (T key), and there you (hopefully) are.

[Crater]

Suggestions for people suffering a couple of problems....

MJ can't handle a launch without spinning out of control sideways within meters in spite of the control part being correctly aligned: Try deleting all of the craft-specific config files in KSP\GameData\MechJeb2\Plugins\PluginData\MechJeb2 - they're called mechjeb_settings_type_*.cfg and there is one for each craft you've flown. Sometimes they seem to have crap data in them, and the defaults work a lot better.

MJ eats all your monoprop when docking / takes forever / fails / never aligns: The usual cause for this is that you are docking in a rotating frame of reference, which means that unless you align your target port along the normal +/- axis relative to its orbit, it will be rotating This is a function of the fact that it is following a curved path around the planet(oid), and that although you are following a parallel curved path, over the course of a single full orbit, you will have rotated through 360 degrees relative to one another. Aligning normal +/- means that the target port will only spin on it's axis, rather than crawling off around the side of the target vessel. If you try to dock with non-aligned port, with a larger vessel that can't flit around very rapidly on its RCS, then you will spend a disproportionate amount of time chasing the docking axis that is constantly moving away from you. Agile craft can still do it, but aligning the ports, or using higher orbits (which have longer periods and therefor lower rotation rates) will make life easier. The docking computer works fine, but if you give it a difficult job, it finds it difficult...

[Starwaster]

If rotational alignment is important to you, use four regular ports on a quad adapter like this. Make sure to roll into proper alignment prior to docking though

Why bother with with that when MJ2 CAN handle port alignment? Excessive docking ports contribute to lag

[Nibb31]

Or use ORDA for docking.

[Ikkinomi]

strange ... After having all conceivable bugs with Mechjeb 2.0.9 (fail rendezvous, docking fail, landing fail etc ...) same with or without the dll sarbian, I went back to 2.0.8 and since mechjeb no problem. Rdz works. except docking autopilot works but often maneuver unnecessarily causing high consumption of resources. But otherwise, it works!

[Dimetime35c]

Okay I downloaded the one from the space port 1.9.8 and it erases all me parts. Nothing loads in any of the columns. Help.

[DChurchill]

Pretty sure 1.9.8 is not going to work. You need the current version which is in the first post of this thread.

[Nibb31]

Version 1.9.8 is deprecated and was designed for 0.18. Many things have changed since then. Do not use it.

[viperwolf]

strange ... After having all conceivable bugs with Mechjeb 2.0.9 (fail rendezvous, docking fail, landing fail etc ...) same with or without the dll sarbian, I went back to 2.0.8 and since mechjeb no problem. Rdz works. except docking autopilot works but often maneuver unnecessarily causing high consumption of resources. But otherwise, it works!

How much have you tested this, i was looking for the last working version.