Jump to content

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


r4m0n

Recommended Posts

I would like to point out that Majiir answered my previous post on this matter by saying it might not all be from MechJeb

http://forum.kerbalspaceprogram.com/showthread.php/42601-Star-map-stuttering-on-rotation

In addition, he gave us an error that appears from the log, that doesn't seem to incriminate MechJeb, but could be amplified by it.

Well, removing MJ2 solves the problem on clean save. Savegame, where MJ2 was used and removed later, the problem persists. Maybe MJ2 writes something in the save config, or somewhere else. I'm just hoping this little bit of info helps.

Link to comment
Share on other sites

A number of people in the past few pages appear to me to be blaming poor attitude stability on mechjeb. My findings are that the old way of building rockets with SAS modules is no longer valid.

I rebuilt all my rockets fresh with 0.21 and yes, the new SAS is very different. I've tried reaction wheels, I've played around with the construction order and I use the RCS aid to get the placement right and eliminate unwanted torque vectors. The problems with large unwieldy rockets seems to be down to KSP itself but the docking problems and problems near minmus are very real. The hohmann to minmus for example, it does the next to last burn fine then starts a 10,000m/s correction that shoots you out of kerbin system and mostly out of fuel.

I'll see if I can come up with concrete, repeatable files over the weekend. At the moment, I'm only using MJ as a node guide - leaving it on full auto isn't reliable.

Link to comment
Share on other sites

@ transcendentape

I haven't rebuilt all of my rockets, and I originally thought the weird behaviour of MJ was due to my Launch Vehicle (which is haphazard, to be honest, but it always worked flawlessly in the previous versions....it still launches fine under manual control, no problem, it's when MJ is controlling it that it acts up. Any of my ships especially seem to act up near the end of a maneuver at around 0.2, 0.1m/s. I've tried my other ships as well aside from my launch vehicle w/payload and even my "normal ship/rocket" designs without the Launch Vehicle are doing the same thing.

I have NEVER used RCS for controlled flight except for docking, so I can't think that I would need it now, especially seeing that you and other folks don't require it while doing the "regular" part of a flight in space or on landing/takeoff. I'm just stumped as to what's causing the problem as I don't THINK my vessel designs would be the issue - I try to think them out as best I can in terms of SAS module location, RCS thruster location, etc...

I've linked the craft file I sent to Duna last night that caused my post in this thread - it was the first flight I'd done since the update of MJ and KSP. Before you ask, it has a lot of parachutes to make the most of slowing down in the Dunan atmosphere, lol. The decoupler module at the rear is for connecting it to a nuclear probe stage to take it to Duna. This stage is built in-line at launch and is attached to the ship via struts and decouplers, so it's solid. Evan after it disconnects in Duna orbit, though, the ship still acts oddly when MJ is in control, especially when I tried to land it at my existing Duna base.

https://docs.google.com/file/d/0ByOO06teP37Ma2J5dUVpUldTZnc/edit?usp=sharing

Edited by darkstrike
Link to comment
Share on other sites

Im experiencing a bug with MechJeb and SAS.

Try the translatron and use "keep vert".

At this point toggling SAS will still work.

Now, check "Keep H/S" and then uncheck it again.

From this point forward neither MechJeb or the player can turn the SAS back on.

Link to comment
Share on other sites

@ transcendentape [snip] The decoupler module at the rear is for connecting it to a nuclear probe stage to take it to Duna. This stage is built in-line at launch and is attached to the ship via struts and decouplers, so it's solid. Evan after it disconnects in Duna orbit, though, the ship still acts oddly when MJ is in control, especially when I tried to land it at my existing Duna base.

https://docs.google.com/file/d/0ByOO06teP37Ma2J5dUVpUldTZnc/edit?usp=sharing

I'm gonna play with your craft in a fresh install of 0.21.1 with mechjeb as the only mod as I use FAR, NP, and some other mods and I don't want to make a craft file that you can't use. However, on first glance, you've got WAY too much SAS torque. By my count, you have a total of 50 for a craft of only 23.2 tons. For comparison, I just made a rebuild of my 270 ton (at LKO) tanker with a total of 16.5 torque. At launch, it's a 1500 ton behemoth and only has 46.5 reaction torque. Again, I haven't played with it yet, but first glance seems to me that you should remove both the ASAS module and the inline reaction wheel.

Link to comment
Share on other sites

the MJ overcorrecting thing, keep disabling reaction wheels until it works. If you slapped a bunch of them on you probly way overdid it, that's just extra weight and you don't need to turn that fast.

If they were to help control a heavier craft, put them on the stages that get ejected.

In other news, MJ failed hard on 3 mun landings in a row. Was still going over 100m/s when the surface came up to find me. Plenty of thrust, if I turned on the extra thrusters it would shut the throttle down (to blast them on full later), the speed/altitude calculation was just plain wrong.

I just had to do it manually. :/

Link to comment
Share on other sites

the MJ overcorrecting thing, keep disabling reaction wheels until it works. If you slapped a bunch of them on you probly way overdid it, that's just extra weight and you don't need to turn that fast.

If they were to help control a heavier craft, put them on the stages that get ejected.

In other news, MJ failed hard on 3 mun landings in a row. Was still going over 100m/s when the surface came up to find me. Plenty of thrust, if I turned on the extra thrusters it would shut the throttle down (to blast them on full later), the speed/altitude calculation was just plain wrong.

I just had to do it manually. :/

I've noticed that land at target isn't working very well. However, once close i've found aborting and hitting 'land anywhere' works.

Link to comment
Share on other sites

That's the thing though Transcendentape, I tried DISABLING the torque in every combination I could think of on all the wheels/pods (including trying JUST having the TINY reaction wheel in the probe core of the nuclear stage active) and MechJeb STILL does the same thing, so I don't think the fact I have too much torque should matter, theoretically, it should just mean my ship can maneuver faster...? I should also mentioned I tried locking the gimbal on the engines too...

I'll try removing them, but if disabling them doesn't seem to make a difference, I don't think removing them would matter...?

Also, @ PrincePapa - link to the part: http://jenkins.mumech.com/job/MechJeb2/lastSuccessfulBuild/artifact/jenkins-MechJeb2-72/MechJeb2-2.0.9.0-72.zip Unzip it to your KSP/Gamedata directory.

Edited by darkstrike
Link to comment
Share on other sites

No problem! :)

As a comment - is there any reason we don't see SAS turn on when MechJeb is under control of a ship anymore? SAS always used to stop the spin/rotation movement of a ship and lock it in a direction, yet, MJ doesn't use that function in this version....could that be a reason why some of our ships are moving around so much...?

Edited by darkstrike
Link to comment
Share on other sites

mechjeb disables the SAS now, and i think thats a MAJOR flaw, with no mechjeb attached, i can fly just about anything with little to no wobble with the new SAS, but i put mechjeb on there and without the SAS most rockets with more than one engine spin out right as i get to the gravity turn.

this needs to be fixed, i figured the new SAS would work well with mechjeb, cure the massive RCS waste at least, but idk maybe they are working on this? and we will see an update soon?

Link to comment
Share on other sites

hellion13, re: ship wobble.

I used to have this problem a lot, and I've learned that it is primarily due to excessive flexibility of the ship structure between the control surfaces and the controlling pod or part ("Control from here"). By stiffening the ship, MechJeb's control inputs become more precise, thereby eliminating wobble.

I'm not sure if this is the issue you're having, but it's made my life a ton easier, and I hope it helps.

However, SAS does have it's purpose and I agree with you that disabling it isn't wise. Unless the update to 0.21 has changed the way it works, my recently acquired understanding of what it actually does - and doesn't - means that it makes good sense for MJ to use it as much as possible. I used to look at the SAS light flickering and call it "MJ's happy light", meaning that when it's on, the steering error is about zero.

If I'm not mistaken, SAS is useful only for stopping the ship's rotation. That tells me that if MJ is rotating to a particular heading and needs to slow rotation as it comes close to it, that is when SAS makes sense, and it seems logical for MJ to use it.

On the other hand, slay_mithos makes a valid point that what SAS is doing to slow the rotation can also be done with control inputs.

The thing is that I've discovered MJ doesn't actually use the entire range of input (I think), meaning it won't push the stick all the way in any direction - I can do more manually than MJ can - which seems like a flaw to me. MJ should have full control and be able to push to all stops, but there have been times when taking manual control, but still in the same direction MJ is trying to go, saves the ship from crashing.

I feel like I'm not making any sense, so if I'm not, I apologize. :huh:

Link to comment
Share on other sites

SAS controls everything now, not just wobble, including heading.

Personally, I'm not seeing any major problems with MJ at the moment (although I haven't tried kethane yet), but I have noticed that MJ is *far* more aggressive in using control forces while keeping a heading than the new SAS is, and that's not really a good thing. I can use SAS now on virtually all of my rockets with no oscillation, but when I utilize MJ for a long interplanetary burn, it sets up a pretty severe oscillation that sometimes damages my craft. (the author) may want to consider using SAS unless SAS isn't being aggressive enough to keep the heading.

Link to comment
Share on other sites

Sure, let me go give it a whirl Sarbian!

*some time later* Ok, just did a test launch of the exact same craft I attached earlier in this thread - launched it from Kerbin to orbit, and then re-landed the ship back at KSC, all with MJ.

The new file IS quite a bit better - launch was a LOT more smooth; my launch vehicle/payload didn't wobble or "chase" the reticle at all. However, on landing, although MJ was smoother, there was still some "reticle chasing" once it got down to 0.2, 0.1m/s during the "Course Correction" stage - a bit less than before at least though...?

Link to comment
Share on other sites

I installed the Version 1.9.8 of MechJeb for a few Weeks. After Update Kerbal Space Programm i want to update the MechJeb Version 2.0.9. I use a Mod Manager and remove the mod Version 1.9.8. Then i installed the Version 2.0.9, but Ingame shows me the MechJeb Version 1.9.8.

Whats happend? Can anybody Help me ?

Link to comment
Share on other sites

Ok, so I see on the latest build that MechJeb is supposed to work with .21... What's the trick to get it working?

EDIT: Ok, nevermind, I figured it out. I can't have KSP installed to my program files folder, something with permissions effects adding mods. I changed my install location for KSP and MechJeb is working. Don't know why this was a issue, but that's what fixed it.

Edited by Wurmfist
Update: Issue Resolved
Link to comment
Share on other sites

Just crashed another keth ground station into Minmus. I was in a 25km circular orbit, picked a kethane spot right round the other side of the planet to give MJ thinking time and set it to auto land. I could see it was going to plant it into the surface but in the interests of science I just sat back and watched. It crashed into the side of a hillock at 154m/s. The prediction was dancing all over the place, is it the new surface contours throwing it off?

Also getting to minmus is causing MJ headaches and it seems to be worse when the Mun is close to the orbital path taken. I'm just plotting my own nodes and letting MJ execute them. Trying rendezvous with a final correction burn seems to end up leaving Kerbin altogether or smacking into Minmus.

Link to comment
Share on other sites

For some reason the ascent guidence system switches between coasting to edge of atmosphere and verticle ascent as soon as my ship reaches 1000 metres. However, this only happens on my 3m ship made of NovaPunch parts whereas my 2metre NP ship works just fine :S

Link to comment
Share on other sites

The next Kethane ground harvester got to minmus and I tried again. From orbit it's very hard to tell what the surface terrain looks like so I just picked what looked flattest with Keth in it and crossed fingers. The unit came down on the side of a valley hitting at about 0.8m/s and survived. From watching MJ I'm sure now that it's blithely unaware of the surface contours.

Better do those landings with a healthy dose of manual ;)

Edit: added screenshots

9C8avMi.jpg

uMO9v4L.jpg

Edited by Ratzap
Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...