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

[sarbian]

Those from MechJeb or the MechJeb Ext are harmless. The mods try to load a something that is not present (the ext check to the dll of the add they support, MJ search for mod implementing MJ class).

The "KineTechAnimationLibrary" one seems more serious. I don't know the addon it comes from and a quick search show a lob from 2013.

You may want to do a clean install of KSP. Some addon may be quite outdated and incompatible with your current version.

[smjjames]

The kinetech thing is from B9 and has already been mentioned on the thread for it, no responses to it. Not sure what it does in B9 though, besides animations obviously.

[Starwaster]

Nope, there is only 1 exception related to MJ in this log (and 2 to the exp) and they only show on init and have no impact on the stability

One of your crash has :

"DynamicHeapAllocator out of memory - Could not get memory for large allocationCould not allocate memory: System out of memory!"

So I guess you are playing near the memory limit. Or you have too many other program running. I managed to have my stripped dev KSP crash of an out of memory with Chrome tabs eating all my mem yesterday...

Yeah, Chrome is pretty hungry about memory when you start abusing the tabs.

I often have so many tabs open that not only do the tabs not display any text, but they don't display icons either. (too many tabs so tabs too small)

What's really bad is that some of those tabs have been there for WEEKS! (usually NASA or NTR stuff)

Not at this moment though. Right now I only have 17 tabs open.

[sarbian]

I always have a minimum of 2 windows full of tabs and often go up to 4. U have this bad habit of oppening a new tab for each search and never closing them. And I end up with 10GB of chrome ....

[futrtrubl]

I was seeing some odd behavior in the Ascent Guidance module. When it switches to "fine tune apoapsis" it turns to prograde instead of staying on the ascent profile. After that for "coasting to edge of atmosphere" it switches back to the ascent profile vector, but it stops obeying "limit angle of attack" which can cause loss of control and with FAR can increase drag considerably.

Just thought it worth mentioning.

Edward

[Starwaster]

I was seeing some odd behavior in the Ascent Guidance module. When it switches to "fine tune apoapsis" it turns to prograde instead of staying on the ascent profile. After that for "coasting to edge of atmosphere" it switches back to the ascent profile vector, but it stops obeying "limit angle of attack" which can cause loss of control and with FAR can increase drag considerably.

Just thought it worth mentioning.

Edward

Hrmmm, I submitted some code awhile back to make it stay on the ascent profile... for the coast phase. The code is in the most recent dev version but I guess I didn't look at the fine tuning phase. Since I put that bit in, I'll take a look at it

(funny, I thought I saw what you were talking about but it was only for a second and then it stopped doing it so I thought I imagined it...)

[Vorg]

No, I'm not Volunteering. It needs to be someone who both knows MJ and is good at documenting. I am nether.

[futrtrubl]

Hrmmm, I submitted some code awhile back to make it stay on the ascent profile... for the coast phase. The code is in the most recent dev version but I guess I didn't look at the fine tuning phase. Since I put that bit in, I'll take a look at it

(funny, I thought I saw what you were talking about but it was only for a second and then it stopped doing it so I thought I imagined it...)

Yup, should have mentioned this was with dev 401. Thank you so much!

[Whirligig Girl]

Do you have a copy of MechJeb 1 for 0.16?

[lodestar]

An issue experienced by some Linux x64 users is FPS drop or stuttering when dragging the camera around with the right mouse button. Yesterday I figured this also happens with the left mouse button and doesn't happen when moving the camera with the keyboard, which strongly suggests some UI event bug, and I further realized it doesn't happen when Mechjeb isn't loaded or when no window is visible. Should I report that as a Mechjeb issue, or there's an Unity UI toolkit being used?

[sarbian]

GregroxMun: The Archaeology department is run by Dr. Jones. Ask him.

lodestar: an interesting one. The only part of MJ that check of the right mouse button is the code to move the menu, and it should not even activate if you click outside the menu button. So I don't see how it could slow down things. But I may forget some other part of the code...

I m working on performance with MJ. When finish that code we will see if it's a MJ or Unity problem.

[Vorg]

Trying some other option in MJ last night to align a polar sat orbit with the one requested by the mission. I had a polar orbit, but started at the wrong time, so looking down it was off by 90 degrees. Looking for a burn to fix it, I would have needed at least twice what I had. In one of the MJ options I found one about moving the AP and PE and it looked like what I wanted, but it request data entered in minutes.seconds where the mission provided the requested orbit in degrees. Any way MJ could get an input format option to select between the two?

[falken]

One thing I would love to see in MJ, is a sort of programmable Smart ASS SURF Mode. I find that SURF mode is more reliable with making an ascent to orbit with FAR than the Ascent guidance (this is with Sarbian's FAR module).

For instance, you could queue up some commands, such as this, which is my NASA style shuttle ascent profile I stick to for MJ with FAR.

Launchpad: Heading 90, Pitch 90, Roll 90

200m: Heading 90, Pitch 70, Roll 180

7500m: Heading 90, Pitch 55, Roll 180

12000m: Heading 90, Pitch 35, Roll 180

22000m: Heading 90, Pitch 5, Roll 180

36000m: Heading 90, Pitch 0, Roll 180

48000m: Heading 90, Pitch 10, Roll 0

I don't know if anyone is really interested in this idea, I just thought I would put it out there.

[futrtrubl]

Trying some other option in MJ last night to align a polar sat orbit with the one requested by the mission. I had a polar orbit, but started at the wrong time, so looking down it was off by 90 degrees. Looking for a burn to fix it, I would have needed at least twice what I had. In one of the MJ options I found one about moving the AP and PE and it looked like what I wanted, but it request data entered in minutes.seconds where the mission provided the requested orbit in degrees. Any way MJ could get an input format option to select between the two?

Changes in inclination are REALLY expensive in dV. A change of 90 degrees like you said is ~1.4 x current orbital velocity, the same as reaching escape velocity. Can't get around that.

I know you said it's more about changing the LAN, but that is equivalent.

Edited February 21, 2015 by futrtrubl

[Pecan]

One thing I would love to see in MJ, is a sort of programmable Smart ASS SURF Mode. I find that SURF mode is more reliable with making an ascent to orbit with FAR than the Ascent guidance (this is with Sarbian's FAR module).

For instance, you could queue up some commands, such as this, which is my NASA style shuttle ascent profile I stick to for MJ with FAR.

Launchpad: Heading 90, Pitch 90, Roll 90

200m: Heading 90, Pitch 70, Roll 180

7500m: Heading 90, Pitch 55, Roll 180

12000m: Heading 90, Pitch 35, Roll 180

22000m: Heading 90, Pitch 5, Roll 180

36000m: Heading 90, Pitch 0, Roll 180

48000m: Heading 90, Pitch 10, Roll 0

I don't know if anyone is really interested in this idea, I just thought I would put it out there.

You might like to look at the KOS scriptable autopilot for that sort of thing.

[Vorg]

The amount of Dv needed isn't the issue. It's that MJ was asking for the target info in minutes.seconds when Kerbal provides the info in degrees. I was asking for an option in MJ to except the data in the same format provided by the stock parts of them game.

[Pecan]

The amount of Dv needed isn't the issue. It's that MJ was asking for the target info in minutes.seconds when Kerbal provides the info in degrees. I was asking for an option in MJ to except the data in the same format provided by the stock parts of them game.

One degree is 60 minutes of arc (and one minute = 60 seconds), so the 'conversion' is really easy.

[armegeddon]

One degree is 60 minutes of arc (and one minute = 60 seconds), so the 'conversion' is really easy.

Or even easier, just type the required degree, with decimal numbers, into the degree field. It ends up close enough that it always works for me.

[sarbian]

falken : I have plant for something related on my list. But first I want to finishing looking at the general MJ performance.

Vorg : you can input a decimal angle in the first input box. it will work.

Edited February 22, 2015 by sarbian

[falken]

That is super, sarbian.

[smjjames]

Any idea what's up with it doing that wierd (and fuel wasting) somersault at the final 500m descent? I've upgraded it from 401 dev version to 402 and it still does it sometimes

Edited February 23, 2015 by smjjames

[smjjames]

Hope you guys don't mind if I make a doublepost.

Uh, I've just had something REALLY odd with mechjeb. For the first time in, well, ever, it missed altitude end point for ascent by 15-16km. The first time, I thought it was some kind of fluke, but when it happened the second time, it did the same. So, I looked at the output logs and there were a bunch of FAR errors (from winglets) which may be responsible, or maybe it's the KW rocketry engine? I'm looking into it. It may not have actually been MJs fault.

Edit: Took off the wings, didn't fix the problems, then replaced the KW rocketry engine with a stock one and bam, fixed.

Edit2: Put the KW rocketry Griffon XX back on and it's doing that behavior again. I haven't had problems with other KW rocketry engines or Novapunch. Not sure if it's a problem with the engine itself or MJ.

Edit3: Tested with another KW engine that I've used before without problems (Titan-T1) and, no problems. I think it may be an issue with the Griffon XX engine, but not being a modder, I have no clue and theres nothing in the logs that would hint at anything.

I would post in the KWrocketry thread, but other than making MJ overshoot the target altitude, I have no data or anything to show a problem.

Edited February 23, 2015 by smjjames

[Neutrinovore]

@smjjames,

Yeah, that's why I don't like it when engines have 'response times' coded into them. Meaning that they take time, usually several seconds in the case of KW Rocketry engines, to spool up or down from zero to full power. There's an alternate config folder in the latest KW archive that lets you go back to having 'instant' response times for the engines, you might try that to see if it solves your problem.

Hope this helps!

[smjjames]

The Griffon XX is apparently the only one that gives MJ issues, for me at least. I'll check out the instant response configs.

Actually, I've noticed that with some engines, when I have a really heavy rocket, it takes a bit of time for it to get up to speed, but I thought that was just because of the weight.

Edited February 23, 2015 by smjjames

[Starwaster]

Any idea what's up with it doing that wierd (and fuel wasting) somersault at the final 500m descent? I've upgraded it from 401 dev version to 402 and it still does it sometimes

It's because when it enters that phase it reorients to prograde which may be different from the flight path it was handling previously.

That in itself isn't a problem. What is a problem is that it's usually using a gimbaled engine for that sudden shift. And then it shuts the engine off because it finished fine tuning its apoapsis. So you have a sudden shift in angular momentum initiated by an engine that is now no longer available to finish the course change leaving you with only RCS and reaction wheels.

Someone else asked that question not long ago and I'll repeat what I said then: I submitted some minor changes to Sarbian that were accepted and they eliminate MOST of what you're describing by having MJ continue following its flight plan instead of shifting orientation to prograde. And that works fairly well except that I missed that one spot during the fine tuning phase.

I'm testing another minor change and if it fixes it and doesn't screw anything up then I'll submit that one too. (I'm not a regular contributor, just the occasional small changes here and there)