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

[AntiSol]

Hi, thanks for mechjeb, it's awesome!

I'm using the latest dev build and I think there's something wrong with the terminal velocity calculations. The 'limit to terminal velocity' option kicks in during ascent but not nearly as early as it used to. Perhaps the new aerodynamics stuff in v1.0 is causing issues?

I've created a custom window to show me terminal velocity and I've kept an eye on it over the course of a few launches, and it doesn't seem to behaving correctly to me.

My understanding is that terminal velocity should be around 110m/s at around 1000m, and should climb as the atmosphere thins during ascent - I remember reading a tutorial ages ago when I first started playing which said that I should be doing 110m/s at 1km altitude and 220m/s at 10km. Previous versions of mechjeb accelerated at about this rate by throttling down due to hitting terminal velocity very early during ascent.

But here's what I see: When my ship is on the pad before I launch, the terminal velocity reads either NaN or ~70m/s. But as soon as I launch (less than 200m from the pad), It reads somewhere around 400m/s and starts falling during ascent. As the terminal velocity figure falls, my ship's speed increases. this goes on until somewhere around 5kms altitude, at which point the ship has accelerated up to about 330m/s and terminal velocity has dropped to 330m/s, so the "limit to terminal velocity" option kicks in and mechjeb throttles down.

After a little while (I think it's around 8km altitude), the terminal velocity speed starts rising again and mechjeb throttles back up and gets me into orbit.

It's been a while since I played kerbal (I heard about the 1.0 release and decided to see what was new), so perhaps something has changed or I'm missing something. I seem to be using more fuel to get into orbit using mechjeb than doing it manually and limiting my speed as mentioned above (220m/s at 10km).

If there are any logs or more detailed info I can provide to help track this down, or tests you'd like me to do, please let me know. I'm running KSP via steam on ubuntu, using the dev build from a few hours ago.

Thanks!

[Snowwy]

What version of MJ since you are using KSP 0.90? Support for an old version is going to be very sketchy.

Are both orbits circular? If either of them somewhat eccentric MJ is going to have difficulties.

What TWR for the craft? Low TWR craft will have trouble as maneuvers are assumed as instant changes of velocity and the less instant they are the more off they are.

Even after all of that you will likely have to "fine tune closest approach" later. SoIs are big, spacecraft aren't.

futrtrubl,

Thanks! I have CKAN installed (mostly so I don't have to track things like that)- the current MJ install is 2.4.2. This has happened with a number of ships with TWR ranging from .1 to 1 (I tend to skimp on space propulsion for the sake of stability- even when I have stronger engines I tend never to run them hard. Yes I know this means I'm building them wrong. ) Most recently a 100-ton-class spaceplane (91 tons after orbital refuelling) with a TWR of .43 missed. The orbit of my Deep Space Launch station is in fact circular, and my intercepts are starting from circular orbits around Kerbin, usually in the 150-300km altitude. Sounds to me like that assumption you mention is in fact the source of the problem, as it takes a fair bit of delta-v to make the climb. If I'm throttled down all the time that would tend to exacerbate the issue, yes?

[AgentMOO]

Re-installing KSP fixed the circularization burns for me.

[Snowwy]

Re-installing KSP fixed the circularization burns for me.

I'll try that immediately.

[AgentMOO]

I'll try that immediately.

And the problem came back on the second launch. =\

*In between the launches I went back to the main menu to change my resolution.

Edited May 12, 2015 by AgentMOO

[Snowwy]

And the problem came back on the second launch. =\

Ouch. I did in fact find that my MJ install was problematical. I'm starting a test launch now and will report back.

[GERULA]

Use the latest dev version.

Dev version, to which smjjames refers, is here: http://jenkins.mumech.com/job/MechJeb2/

thanks guys

[futrtrubl]

futrtrubl,

Thanks! I have CKAN installed (mostly so I don't have to track things like that)- the current MJ install is 2.4.2. This has happened with a number of ships with TWR ranging from .1 to 1 (I tend to skimp on space propulsion for the sake of stability- even when I have stronger engines I tend never to run them hard. Yes I know this means I'm building them wrong. ) Most recently a 100-ton-class spaceplane (91 tons after orbital refuelling) with a TWR of .43 missed. The orbit of my Deep Space Launch station is in fact circular, and my intercepts are starting from circular orbits around Kerbin, usually in the 150-300km altitude. Sounds to me like that assumption you mention is in fact the source of the problem, as it takes a fair bit of delta-v to make the climb. If I'm throttled down all the time that would tend to exacerbate the issue, yes?

All I can suggest is to try a very simple craft with high TWR, plenty of reaction wheel control and no flex and try a test rendezvous. At a TWR of 1 the burn shouldn't be longer than 10 seconds and at a start alt of 150-300km that should be close enough to instantaneous. See what your closest approach is then, and how much dV is required to "fine tune closest approach".

Oh and before burning to rendezvous "match planes with target" of course so we don't have that complicating things.

[Snowwy]

All I can suggest is to try a very simple craft with high TWR, plenty of reaction wheel control and no flex and try a test rendezvous. At a TWR of 1 the burn shouldn't be longer than 10 seconds and at a start alt of 150-300km that should be close enough to instantaneous. See what your closest approach is then, and how much dV is required to "fine tune closest approach".

Oh and before burning to rendezvous "match planes with target" of course so we don't have that complicating things.

Whoa! That's CRAZY fast, to me. I better throw some big stonking reaction wheels on my next flight and see what it does.

I can report that my current test flight (much lighter 40-ton spaceplane with starting orbital TWR of .8) did in fact perform an accurate intercept burn (thank you for your help!). I kept it choked off as always so my burn took more like two minutes, but instead of throwing me clean out of Kerbin's SoI it instead gave me a 200km intercept. That's completely acceptable and your input goes a long way to explaining what was going wrong. I'll try it with the heavy next, but for now I'm calling it solved.

[futrtrubl]

Whoa! That's CRAZY fast, to me. I better throw some big stonking reaction wheels on my next flight and see what it does.

I can report that my current test flight (much lighter 40-ton spaceplane with starting orbital TWR of .8) did in fact perform an accurate intercept burn (thank you for your help!). I kept it choked off as always so my burn took more like two minutes, but instead of throwing me clean out of Kerbin's SoI it instead gave me a 200km intercept. That's completely acceptable and your input goes a long way to explaining what was going wrong. I'll try it with the heavy next, but for now I'm calling it solved.

That IS crazy fast ;'] I would normally never have a purely orbital craft/stage with that much TWR as the inefficiency of a long burn is usually far outweighed by the inefficiency of lugging extra engine mass along.

Having said that, if you are lugging extra engines around anyway it's usually better to use them, unless of course they are lower ISP or cause control or structural problems ;']

This is just to test if what I think is the issue actually is and it isn't a bug in MJ.

[sarbian]

I have CKAN installed (mostly so I don't have to track things like that)- the current MJ install is 2.4.2.

I just want to point out that the dev version of MJ are in CKAN. You just have to go in the option and add the MJ Dev repository (it s in the default list).

[Kitspace]

Wait does that mean it is possible to have an automatic install of a development build?

[sarbian]

Yes, my build server post each dev release on CKAN.

[AgentMOO]

Any theories why MJ stops doing final circularization burn with ascent guidance? I've rebooted, re-installed, and have no mods other than MJ.

It's not my craft. Here is the file. (http://www./download/9cyct9bnp4hrh76/newb5.craft)

And here is a picture of my settings.

Edited May 12, 2015 by AgentMOO

[tmikesecrist3]

it cant plot the cerulization burn if you dont have your tracking center upgraded

[Shadriss]

it cant plot the cerulization burn if you dont have your tracking center upgraded

While true, look at his screenshot. The Tracking Center can be clearly seen, and it's fully upgraded.

I'm actually seeing this same problem. On Acent w/ autopilot, it will autostage and autocorrect course, but there is no throttle control whatsoever. As a result, once you do hit the target apoapsis, the ship just keeps right on going. Had no issues until about five or six days ago. Tried the latest build (21 hours ago, as of this post) and tried it again, and throttle still seems to be untouched.

[sarbian]

It does look upgraded on that screenshot but it would indeed be a good answer.

[AgentMOO]

it cant plot the cerulization burn if you dont have your tracking center upgraded

It's in sandbox mode. I flew many flights yesterday it was working fine. Then it just stopped working.

[SaintWacko]

Even wihout Persistent Rotation, the stock game does that.

I've noticed it does sometimes, but only very rarely, and not by much. With Persistent Rotation, it always rotates away from the node during warp, and by quite a lot.

[smjjames]

Hm, build 455, I got a bunch of arthrimetric exceptions when my craft splashed down, none of the autopilots was active, although I did timewarp almost all the way down from the point that I opened my chutes. However, while those exceptions briefly spamming in the exception detector caught my attention, theres some other exceptions that I noticed which went earlier but didn't show up in the exception detector.

Edit: Did some quick testing (warping from chute deploy to almost splashed down and one without warp) on that splashdown error spam to see if it was reproducible or a one off and it happened both times. An observation though, it doesn't happen right away, it happens when the craft tips over in the water. Actually, I have an idea on a third test.

Edit2: Did a quick third test by having SAS on so that it didn't tip over soon after splashdown. I let it splashdown with SAS on, waited a couple seconds (5 or 10 seconds), turned off SAS, then it slowly tipped over, no exception spam.

Output log: http://sta.sh/02es9m9btajq

The exception that didn't appear in the exception detector is this one:

MechJeb module MechJebModuleThrustController threw an exception in Drive: System.InvalidOperationException: Operation is not valid due to the current state of the object

  at System.Linq.Enumerable.Iterate[Part,Double] (IEnumerable`1 source, Double initValue, System.Func`3 selector) [0x00000] in <filename unknown>:0 

  at System.Linq.Enumerable.Max[Part] (IEnumerable`1 source, System.Func`2 selector) [0x00000] in <filename unknown>:0 

  at MuMech.MechJebModuleThrustController.TemperatureSafetyThrottle () [0x00000] in <filename unknown>:0 

  at MuMech.MechJebModuleThrustController.Drive (.FlightCtrlState s) [0x00000] in <filename unknown>:0 

  at MuMech.MechJebCore.Drive (.FlightCtrlState s) [0x00000] in <filename unknown>:0 

It seems to have triggered right when I smashed a probe into the ground (manually, not on autopilot). I saw the same thing in the output log yesterday, but the log was so bloated (almost 230mb) with messages from the interstellar fuel switch plugin (Freethinker is working on that issue) that my attempts to remove those lines just made notepad++ freeze up, so I decided to spare you from having to deal with that and just report it the next time I saw.

Anyway, is the 'threw an exeption in drive' from MJ something that I can ignore?

- - - Updated - - -

I've noticed it does sometimes, but only very rarely, and not by much. With Persistent Rotation, it always rotates away from the node during warp, and by quite a lot.

I've seen it happen while orbiting small moons (the asteroidal moons especially) and occasionally in low altitude orbits, it can happen with highly eccentric orbits as well.

Edited May 12, 2015 by smjjames

[AgentMOO]

While true, look at his screenshot. The Tracking Center can be clearly seen, and it's fully upgraded.

I'm actually seeing this same problem. On Acent w/ autopilot, it will autostage and autocorrect course, but there is no throttle control whatsoever. As a result, once you do hit the target apoapsis, the ship just keeps right on going. Had no issues until about five or six days ago. Tried the latest build (21 hours ago, as of this post) and tried it again, and throttle still seems to be untouched.

Yes and it doesn't auto-warp either. I am just piloting the node manually for now. Then I remove the node with the maneuver planner after and MJ starts working fine again for other actions.

Here's another pic of exactly when it ceases to function.

[smjjames]

Yes and it doesn't auto-warp either. I am just piloting the node manually for now. Then I remove the node with the maneuver planner after and MJ starts working fine again for other actions.

Here's another pic of exactly when it ceases to function.

http://i.imgur.com/YhsS3O1.jpg

While not MJ related, I'm just letting you know that there is a memory leak associated with those temp gauges (F10), so you're going to want to turn those off. Theres a few plugins around that forces it off so that you don't have to press F10 every time to go to a new craft, or do a scene change I think.

I question your turn shape there, but then again, I haven't flown a spaceplane, so what do I know anyway.

[AgentMOO]

Good to know. Thanks, smjjames. As for the turn shape. It was something I figured out along time ago. I should probably experiment with it again. The idea is to pick up as much speed in the atmo as I can as well as minimize the time it takes to get to orbit. I make lots of revisions to my designs so the faster I can test them the better. Also, making course corrections at those speeds in atmo can rip your ship apart or slow it down.

[AgentMOO]

OK so maybe it did have something to do with my craft... I got around to testing different engines I removed the electric powered ones ( I wasn't using them during ascent ) and all the nuke reactors from within the ship that powered them. Replaced the crew compartment/cargo bay with another LF/O tank, and added a couple THUD engines to the rear. I don't know why but MJ started doing the circularization burns again. I've only done one test so far.

http://www./download/gn382p8r24t5w3o/newb7.craft

Could it be something with Electric engine? Or maybe having to many power-plants?

Update: Now it's not doing the circularization burn for the rendezvous autopilot. Strange mojo...

Edited May 12, 2015 by AgentMOO

[Oron]

I see people talking about the ascent guidance, I'm not to sure if its the same issue as mine.

However, I can never seem to be able to get into orbit with the ascent guidance, once in space ascent guidance turns itself off and just sits there.

I could be doing something wrong and its completely my fault but I've tried so many times, I even used the dev build and still the same thing.

Am I missing something, should I be doing anything different to have Mechjeb take me into orbit?

I've looked at videos, some wore very old and unhelpful, others just show that pressing ascent guidance was all you need to do and bam, your in orbit, not in my case.

My apologies if this has been asked, I tried to search for the problem I am getting with ascent guidance, but did not find anything.

If anyone can help me out, thank you.

Edited May 12, 2015 by Oron