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

[Starwaster]

Ok, took another look at this taking into account what MOARdV said and it's definitely not a case of not having a control reference set. When it happens, setting 'control from here' doesn't fix it. Two craft of identical mass were docked and when I undocked them one would consistently set its Tf to 1.0 even though it was a 2 ton craft. (typically before docking they would set Tf to 0.04 which IMO is too low but that's what they would set it to BEFORE docking & undocking)

Also while (re)testing this I also noticed then when I undocked craft 1 and craft 2, sometimes while setting SmartASS commands for Craft 1, Craft 2's RCS thrusters would sometimes start firing. Typically this would be KILL ROT and Craft 2 would just start firing/counterfiring repeatedly. Keep in mind this is while setting commands for Craft 1

Now... even while submitting this report/feedback and also keep in mind that I do feel docking/RCS in general do need a good bit more work, I want to take this opportunity to showcase a (IMO) nearly impossible task that I set MJ2 to perform. For those who don't know, I've been working on a set of parts to re-create in KSP the Copernicus MTV (Mars Transfer Vehicle). And possibly the Terra Nova from Race to Mars if you've ever seen that. One thing that I wanted to make possible is the re-docking of the drop tank so that the craft could be re-usable, even though the NASA proposal calls for the thing to be dumped in heliocentric orbit. (what... really NASA?)

I did not think MJ2 could perform this task given its tendency to approach the target at something like a 45 degree angle and I meant to try tweaking this before attempting this task but somehow forgot to actually apply the fix that I wanted to try. (dunno why, it's not like it was mindbogglingly hard but I just forgot). So, anyway... here it is. Tonight, it did it. It bumped the tank a few times and nearly threw me into cardiac arrest but it somehow did it. And, even when it lines up successfully and moves forward, there's no guarantee that it can dock the thing because the truss is not a convex shape so it can't actually contain the tank! Even so.... MJ2 DID THIS

(Edit: So, when someone comes in here and complains about how crappy the docking AP is and how it smashed their station to bits, just direct them to this post. RCS on this thing was a bit off balance because of the mass of fuel in the back, the engines and the fact that it didn't fully line up with the truss so the RCS thrusters were a few degrees out of alignment)

Javascript is disabled. View full album

Edited January 17, 2014 by Starwaster

[Newbinator]

Since updating MJ and I have also rolled it back a couple of versions, I have noticed the auto-pilot landing is completely broken (it may be me) but when you select a place to land and select land here, the craft does a 306 degree turn and then goes off on a different tangent.

I dunno if this caused by other mods but I am wondering if anyone has had similar problems and if so, how did you fix it.

Appreciation

Newb

[Castun]

With the latest version, I had a couple custom windows setup, one being for custom rendezvous information. I noticed a bit of lag, and checked out the debug log, turns out it was spamming the log with NaN errors until I deleted the custom windows.

[AbeS]

Since updating MJ and I have also rolled it back a couple of versions, I have noticed the auto-pilot landing is completely broken (it may be me) but when you select a place to land and select land here, the craft does a 306 degree turn and then goes off on a different tangent.

I dunno if this caused by other mods but I am wondering if anyone has had similar problems and if so, how did you fix it.

Appreciation

Newb

are you using FAR?

[Newbinator]

No...I have just reinstalled everything and will continue testing. I thought it may have had something to do with the compressor for textures but alas its not

[Starwaster]

Since updating MJ and I have also rolled it back a couple of versions, I have noticed the auto-pilot landing is completely broken (it may be me) but when you select a place to land and select land here, the craft does a 306 degree turn and then goes off on a different tangent.

I dunno if this caused by other mods but I am wondering if anyone has had similar problems and if so, how did you fix it.

Appreciation

Newb

are you using FAR?

It's not a FAR issue. Either he's controlling from a pod/port that's off-axis or he's run afoul of a long standing bug where the re-entry controller will (usually after an initially successful de-orbit burn) start doing 'course corrections' that take it off course. And it continues to perform the burn until it runs out of fuel. (even to the point where it reaches escape trajectory. I once turned on fuel cheats during such an event just so I could see when it stopped and I tired of the exercise long before MJ2 did)

Sooo, Newbinator, make sure that you're controlling from a forward facing command pod, probe core or docking port and try again. It's either that or the bug I described and if it's the bug, it usually stops after you reload. Or try exiting to spaceport and try again if it persists, but it's never hit me more than once. YMMV.

[Newbinator]

It's not a FAR issue. Either he's controlling from a pod/port that's off-axis or he's run afoul of a long standing bug where the re-entry controller will (usually after an initially successful de-orbit burn) start doing 'course corrections' that take it off course. And it continues to perform the burn until it runs out of fuel. (even to the point where it reaches escape trajectory. I once turned on fuel cheats during such an event just so I could see when it stopped and I tired of the exercise long before MJ2 did)

Sooo, Newbinator, make sure that you're controlling from a forward facing command pod, probe core or docking port and try again. It's either that or the bug I described and if it's the bug, it usually stops after you reload. Or try exiting to spaceport and try again if it persists, but it's never hit me more than once. YMMV.

Awesome Help guys,

I remember my craft was facing rearwards (away from target). I will reload and try it again. I will post results..

Thanks again

Edit

I think I have narrowed it down to the Alternis Kerbol mod.

I used another copy of KSP without that mod (the rest were kept as a benchmark) and it seem to do the right thing. Landed exactly where it was supposed to and NO millions of course corrections.

Pity I really wanted to play that mod.

Edited January 17, 2014 by Newbinator
Troubleshooting

[Gnipu]

Suggestion (is that OK?):

For the Rover window, how about adding checkpoints? I.e. first go to X, then Y. Maybe even different speeds?

Main use would be to avoid terrain.

[codepoet]

Awesome Help guys,

I remember my craft was facing rearwards (away from target). I will reload and try it again. I will post results..

Thanks again

Edit

I think I have narrowed it down to the Alternis Kerbol mod.

I used another copy of KSP without that mod (the rest were kept as a benchmark) and it seem to do the right thing. Landed exactly where it was supposed to and NO millions of course corrections.

Pity I really wanted to play that mod.

Does that mode have significantly differently scaled bodies or bodies with unusual atmosphere characteristics? It might well be that the procedure MJ follows does not work for bodies with particular characteristics that do not exist in the stock game. Does anyone have experience of landing on bodies that are scaled to "real" size - is that the RSS mod or somethign similarly named? I have not really looked at those sorts of mods myself.

[sarbian]

Starwaster : MOARdV is not far from the reality. Something is not set properly after an undocking/decoupling. I added some code to avoid filling the log with error. The problem is I still don't know if it's a problem within MJ ( An error with a callback as a.g. suggested to me ) or a problem with .23 itself. I need to write a small debugging addon to test that theory.

Newbinator : as Starwaster said here a long standing bug I have yet to find/fix. A savegame and instruction to duplicate the bug would help.

Castun : Can you check the log in KSP_DATA and paste the whole error. It's easier to fix if I don't have to spend time trying to duplicate.

Gnipu : The waypoints system for the Rover AP is in since mid December. It's even in release v2.1.1.

And the news : I have been working on the RCS/Docking AP for the last few days. The docking AP logic seems to be OK but I found a bug with the RCS logic, at least I think I found one.

There is a call to get how much thrust the RCS system can give you in a given direction. When I test it and ask how much thrust I have to go forward I don't get the expected result.

So either my test is wrong (I don't think so this time) or the function is wrong (and I can't see why yet). Most likely something wrong with coordinate system change.

The good news is that once I fix that RCS use should improve.

Edited January 17, 2014 by sarbian

[crazyewok]

Didnt get a awnser but why does mecjeb flip by rocket in to a 360 degree summersalt when it just needs to tilt slightly the otherway?

[sarbian]

All the 360° loop "problems" I know of are user related :

- You use FAR and turn at high speed low in atmo => drag/lift push the rocket off its path, t does a loop and MJ tries its best to regain control and succeed most of the time since the loop slowed the rocket.

- You don't use FAR but your design is top heavy and do a loop after you dropped all your nice heavy booster engine => CoM move far up and the start of the turn ends badly. Drag may play a part too

Long story short : you need to adjust the ascent profile

[crazyewok]

Long story short : you need to adjust the ascent profile

Thanks.

My rockets are top heavy and I use FAR so I guessing I will need to do that.

[Starwaster]

Starwaster : MOARdV is not far from the reality. Something is not set properly after an undocking/decoupling. I added some code to avoid filling the log with error. The problem is I still don't know if it's a problem within MJ ( An error with a callback as a.g. suggested to me ) or a problem with .23 itself. I need to write a small debugging addon to test that theory.

When a craft other than the one that I set RCS on and send SmartASS commands to starts responding by firing RCS thrusters, doesn't that point towards MJ?

PID values also all set to (-)Infinity. The only other time I've seen that is if RCS is missing from the craft. (which seems odd in itself given that it still has more than enough torque from its flywheels.... were you aware of that one)

Newbinator : as Starwaster said here a long standing bug I have yet to find/fix. A savegame and instruction to duplicate the bug would help.

Just to make sure, you did see this, didn't you? This doesn't help? https://www.dropbox.com/sh/9pxl3a3fz3tetcc/rxV0aqOzPz

It's got logs, screenshots and a save file. No craft file but the craft were all stock. As to instructions, nothing special needs to be done, if you try initiating auto-land (at target) enough it will happen. Happened to me on my second try.

Edit: Nope I take that back. Craft file was included. It's the whole save folder and the craft is in the VAB. I think I used Hyperedit, so not 100% stock but that was just so I didn't have to waste time getting the thing into orbit first. The save file has multiple craft around several planets so I could test as to whether one planet did this more than another.

Edited January 17, 2014 by Starwaster

[sarbian]

When a craft other than the one that I set RCS on and send SmartASS commands to starts responding by firing RCS thrusters, doesn't that point towards MJ?

PID values also all set to (-)Infinity. The only other time I've seen that is if RCS is missing from the craft. (which seems odd in itself given that it still has more than enough torque from its flywheels.... were you aware of that one)

You read/or I wrote my sentence wrong. I know there is a problem in MJ but I don't know if the source of that problem comes from a change in .23 that MJ handle badly, or a true bug in .23

I don't think I saw that one, I'll have a look. Thanks and sorry that I missed it.

Edit : I forgot to mention a patch by bk2w I merged yesterday and that let you select the current body for the Delta V stats in the Editor.

Edit 2 : I just did some test and it works with my version. The errors in the screenshots looks like the bug I fixed in dev #158

Edited January 17, 2014 by sarbian

[Starwaster]

Does that mode have significantly differently scaled bodies or bodies with unusual atmosphere characteristics? It might well be that the procedure MJ follows does not work for bodies with particular characteristics that do not exist in the stock game. Does anyone have experience of landing on bodies that are scaled to "real" size - is that the RSS mod or somethign similarly named? I have not really looked at those sorts of mods myself.

Sorry I didn't see this before. I've used RSS a lot and I didn't experience the behavior that Newbinator did any more than I did in stock. If it wasn't the control point issue and it was the previously discussed bug AND he was using something older than 158 (I think that's where Sarbian said he fixed it?) then IMO it was something intermittent. Removing Alternis and 'Landing AP working' is likely coincidence. Any time it hit me before I'd just F9 and try again.

[Damaske]

I have been noticing a recent roll to one side on my rockets under ascent autopilot, And I've also noticed that Mechjeb does NOT control the ships roll at all. I feel that there need to be a lock to keep the roll locked in a certain potion, ( user selectable) so your able to keep the wings level to help provide lift during space plane ascents.

I've had space planes roll 90 degrees from wing level and then spin out of control due to mechjeb not keeping the wings level. Is there something that can be done aboot this or is it a "stock" issue and unfixable without many headaches and code cracking?

Was asked a few pages back but i don't see any response to it maybe my post has been overlooked?

[sarbian]

It's not being ignored, it's just that I answered a similar question a few page earlier. The ship roll because the ascent AP want to use a specific heading, and if the "control from here" part is not turned that way then you roll on start.

I want to fix it but I need to fix an other one before. In the mean time check that your ship is actualy controled from the pod/probe and not from an MJ AR202 on the side.

[Damaske]

It's not being ignored, it's just that I answered a similar question a few page earlier. The ship roll because the ascent AP want to use a specific heading, and if the "control from here" part is not turned that way then you roll on start.

I want to fix it but I need to fix an other one before. In the mean time check that your ship is actualy controled from the pod/probe and not from an MJ AR202 on the side.

Ah ok, I normally have it on the topside of the craft and not on the side's. and then i do make sure the pod itself is set to control from here. thanks for the response its a lot of pages to read though and i don't mean to sound like a dick.

[Galane]

If the navball isn't showing all blue when the rocket is on the pad, the launch will tend to go bad. Change which part has control from here set.

Could add that to some other old sayings. Red sky at night, sailor's delight. Red sky at morning, sailor take warning. Blue ball on the pad or launch will go bad.

[Commissar]

i seem to have problems occasionally, where the docking autopilot just wants to hover, and eats up all my RCS fuel. it's happened several times... the latest, well, it ate 3000 mono prop. I really don't want to try to dock a 100t tanker with my stations manually...

edit: it turns out that i didn't have the SAS turned on in the station. maybe mechjeb can include a warning if possible?

Edited January 18, 2014 by Commissar
i got it to dock

[Castun]

With the latest version, I had a couple custom windows setup, one being for custom rendezvous information. I noticed a bit of lag, and checked out the debug log, turns out it was spamming the log with NaN errors until I deleted the custom windows.

Castun : Can you check the log in KSP_DATA and paste the whole error. It's easier to fix if I don't have to spend time trying to duplicate.

The specific error was:

[LOG 18:27:37.317] getObtAtUT result is NaN! UT: 19254.37889324

I googled it in relation to KSP and came upon another post from a while back that had the same problem and solved it by deleting a couple windows. This seemed to fix it for me also, but now that I'm trying to recreate the problem, I'm having difficulty. I know the Vessel Info window had a couple NaN values shown instead of 0 or N/a when I was trying to just test out a cockpit by itself.

Also seeing this alot, though I doubt this is from MJ:

[LOG 23:00:51.102] CerberusEngineController.FixedUpdate
[LOG 23:00:51.108] CerberusEngineController.FixedUpdate
[LOG 23:00:51.113] CerberusEngineController.FixedUpdate
[LOG 23:00:51.128] CerberusEngineController.Update
[LOG 23:00:51.132] CerberusEngineController.LateUpdate

[Plaster]

So i have a small space station where i want to get some solarpanels up but when i want to launch the ap is moving around all over the map and for some reason the winglets are also turning all this happens when i engage autopilot and then it goes right in to the ground can anybody tell what might be wrong? (this did not happen when i sent my first core up to the space station) and then i found out that ist the periapsis thats broken is says that ap is 80 km and periapsis is -598.442 km or -598.443 km and the winglets is still turning?

[Damaske]

If the navball isn't showing all blue when the rocket is on the pad, the launch will tend to go bad. Change which part has control from here set.

Could add that to some other old sayings. Red sky at night, sailor's delight. Red sky at morning, sailor take warning. Blue ball on the pad or launch will go bad.

would that also count for space planes? As i have more trouble with space planes over rockets. And i tend to use space planes a lot more than rockets lately.

[AbeS]

with planes launching from the runway, the heading should be 90 deg with blue on top and orange on the bottom