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

[smjjames]

Build 438, staging calculation is still kind of funky.

Correct (those are LFOX boosters btw, not SRBs, and they aren't linked by fuel lines)

Add one more tank to the bottom stage and it screws up.

After a few more removals and adds, the setup in the first image decided to be broken.

No exceptions spotted in the output log.

Edit: MJ threw a bunch of math errors when I launched the ship (second pics setup): http://sta.sh/013obtdv79gh

Edited April 29, 2015 by smjjames

[slaintemaith]

Landing Guidance -> Land At Target still appears to be broken with build #438. This is with Vanilla install, sandbox game. Again, I have no idea where any logs you'd like to view are located. Sorry. I'm not deliberately trying to be ignorant; I simply am. =)

The requested file lives here https://www.dropbox.com/s/hhozdbr0rx7qsx9/output_log_landing.txt?dl=0

Edited April 30, 2015 by slaintemaith

[smjjames]

Landing Guidance -> Land At Target still appears to be broken with build #438. This is with Vanilla install, sandbox game. Again, I have no idea where any logs you'd like to view are located. Sorry. I'm not deliberately trying to be ignorant; I simply am. =)

Go to KSPs main directory>KSP_Data folder> Inside there is the output log which he'd like to view.

I can also confirm that Landing Guidance>Land at Target is still broken (for atmosphere at least). Try doing a polar orbit at around 200km and land at either ice cap.

Edit: You know, it seems to think that the trajectory is a whole lot steeper than it actually is.

Edited April 29, 2015 by smjjames

[Railgunner2160]

Go to KSPs main directory>KSP_Data folder> Inside there is the output log which he'd like to view.

I can also confirm that Landing Guidance>Land at Target is still broken (for atmosphere at least). Try doing a polar orbit at around 200km and land at either ice cap.

Edit: You know, it seems to think that the trajectory is a whole lot steeper than it actually is.

I agree, for whatever reason for atmospheres it seems to switch quite rapidly between the actual current orbit, and some incredibly steep orbit. It also tends to oscillate constantly trying to find the right point for the course correction burn.......

The DV value seems change quite severely in the other direction every few seconds, my craft keeps swinging trying to adjust the course. Probably a symptom of whatever issue is causing the orbit course switch.....

Edit: (This was on both the last version & the current version 438.....)

Edited April 30, 2015 by Railgunner2160

[GeneCash]

Here's some detail on my issues with land at target.

This is stock 1.0 except for MechJeb 2.5.0.0-438 & Hyperedit

Works fine on airless bodies such as the Mun or Minmus, no matter what the ship. I get landings within 15m of target.

Works pretty well on Kerbin with the Mechjeb capsule. Lands within about 140m of launch pad target. Does not seem to work with stock parts.

I made a test ship with the 3-man capsule+Mechjeb module on an X200-16 tank, with the 2.5m heatshield, 3 Mk-55 radial engines, and 3 LT-2 landing legs. (see image)

I launch or hyperedit into a 100km orbit.

Set it to land at KSC pad. Hit "Land at target" button.

While it burns, it's predicting far, far, short of the actual landing point, maybe 20% of the actual distance.

If you turn off autoland, the prediction pops to something of the rough order of magnitude and you can use retro/prograde burns to manually adjust to the usual 150m limit.

However, this prediction is still slightly long of the actual landing, as you'll end up about 3-5km short of KSC if you let it coast in.

Hopefully some images will help

Javascript is disabled. View full album

[chicknblender]

I believe this is a bug:

The Delta-V Stats for designs with drop tanks and fuel lines don't correctly separate the stats for before and after the tanks are empty. Example: MechJeb calculates the delta v for this rocket as if all the fuel will be burned in Stage 1 before the tanks are dropped, and therefore shows no delta v for Stage 0.

Thank you to all of the MechJeb developers! You guys are incredible.

[smjjames]

Theres also some bugs going on with the external fuel lines, that could contribute to it.

[Nothalogh]

It appears that SAS is not able to lock on to the ascent guidance target vector.

Has anyone else noticed this?

[geo950]

oops.

Edited April 30, 2015 by geo950

[futrtrubl]

To those having trouble with loosing control on ascents: That's the new stock aero. You need to make properly aerodynamic launchers probably with fins at the base. You will now need to limit your AoA or you WILL loose control and you need to now do proper gravity turns.

To those missing the mark when landing on bodies with atmosphere: That's the new stock aero. Your drag and lift now greatly depends on your orientation and a host of other stuff. This makes it very difficult to predict, not impossible. What makes it nigh impossible is that Squad didn't release how exactly the stock aero does that so MJ can't predict it. They didn't with the old aero either, but it was very easy as it was just the weighted average of the drag coefs of the parts.

http://www.omfg.ch/ksp/ascent.jpg

THIS won't work. Can someone post his settings?

thank you!

HOW does it not work? There is not enough information here to help you.

[smjjames]

To those missing the mark when landing on bodies with atmosphere: That's the new stock aero. Your drag and lift now greatly depends on your orientation and a host of other stuff. This makes it very difficult to predict, not impossible. What makes it nigh impossible is that Squad didn't release how exactly the stock aero does that so MJ can't predict it. They didn't with the old aero either, but it was very easy as it was just the weighted average of the drag coefs of the parts.

I'm not talking about off by a few hundred meters, I fully expect it to be off, we're talking about many kilometers plus screwing up the landing proccess before it even enters the atmosphere in the first place.

[Tewkes]

Probably a question best asked in the MechJeb thread

http://forum.kerbalspaceprogram.com/threads/12384-PART-1-0-Anatid-Robotics-MuMech-MechJeb-Autopilot-v2-5

Good idea. Merged.

Edited April 30, 2015 by Vanamonde

[Railgunner2160]

I'm not talking about off by a few hundred meters, I fully expect it to be off, we're talking about many kilometers plus screwing up the landing proccess before it even enters the atmosphere in the first place.

Exactly!!! If the landing autopilot can't even put you on the correct course without burning every last drop of fuel in your tanks AND still leave you with a 20km periapse on the opposite side of the planet from your target then there is clearly an issue!!!!!

[Mishkin_007]

cant even get a proper circularization. going to have to sadly abandon this mod until is working properly again.

[stildawn]

Hi all

Firstly thanks again for the awesome mod. Used it back in. 24 and again now that I'm back into KSP after the 1.0 release.

So anyway I don't know if the issue I'm having is a bug with the mod or just the new game version etc but basically I have a rocket that works and is stable when I fly it myself, however when using mechjeb accent pilot, well for some reason it does a flip (a complete flip) when it starts the turn (gravity turn? The first turn right after just accelerating upwards). What is more odd though is that more often than not after its random complete flip it settles back on course and continues on as if nothing has happened. Sometimes it will do two flips then go back on course, and very rarely it will just flip uncontrollably until I revert back to launch pad.

It's an epic waste of fuel these flips, and often messes up the circle burn as well because of it.

So any ideas on what's going on? I find it very odd that this happens randomly even when I'm using the same rocket each time (in this case my standard mun lander rocket).

Also the inclination thing seems very odd, I took some "put satellite on this orbit" missions, set the correct inclination in mechjeb, but the resultant orbit is wildly different to the goals orbit, often times the opposite. I had tried doing a bunch of maths to figure it out but no luck.

Anyway cheers for any help with this, can provide rocket to test etc if required but I might just be doing something wrong.

Edit: just read that others are having issues also, although they mention losing control, my issue is hardly losing control the flips look perfect as in that they are done on purpose and the fact that it seems to easily put the rocket right again 9 out of 10 times seems odd to me.

Edited April 30, 2015 by stildawn

[diegzumillo]

Hi all

Firstly thanks again for the awesome mod. Used it back in. 24 and again now that I'm back into KSP after the 1.0 release.

So anyway I don't know if the issue I'm having is a bug with the mod or just the new game version etc but basically I have a rocket that works and is stable when I fly it myself, however when using mechjeb accent pilot, well for some reason it does a flip (a complete flip) when it starts the turn (gravity turn? The first turn right after just accelerating upwards). What is more odd though is that more often than not after its random complete flip it settles back on course and continues on as if nothing has happened. Sometimes it will do two flips then go back on course, and very rarely it will just flip uncontrollably until I revert back to launch pad.

It's an epic waste of fuel these flips, and often messes up the circle burn as well because of it.

So any ideas on what's going on? I find it very odd that this happens randomly even when I'm using the same rocket each time (in this case my standard mun lander rocket).

Also the inclination thing seems very odd, I took some "put satellite on this orbit" missions, set the correct inclination in mechjeb, but the resultant orbit is wildly different to the goals orbit, often times the opposite. I had tried doing a bunch of maths to figure it out but no luck.

Anyway cheers for any help with this, can provide rocket to test etc if required but I might just be doing something wrong.

Edit: just read that others are having issues also, although they mention losing control, my issue is hardly losing control the flips look perfect as in that they are done on purpose and the fact that it seems to easily put the rocket right again 9 out of 10 times seems odd to me.

You usually have to tweak the ascent curve to fit your rocket. Some rockets require a smoother ascent profile, and that seems to be the case with yours.

[magnemoe]

I'm not talking about off by a few hundred meters, I fully expect it to be off, we're talking about many kilometers plus screwing up the landing proccess before it even enters the atmosphere in the first place.

Atmosphere is totally different, you will be off my many km. Best option is probably to use it then see how much its off and correct, however mechjeb landing autopilot does an pretty fast dive then landing so you will heat up a lot too.

Note if you overshoot you can pop parachute to slow down.

Personally I have problems hitting the desert west of KSC

Landing on Mun works perfectly. Same does all orbital operations.

I would fixing the dV calculation bug first then accent adjustments before looking at landing who will be hard.

- - - Updated - - -

I believe this is a bug:

The Delta-V Stats for designs with drop tanks and fuel lines don't correctly separate the stats for before and after the tanks are empty. Example: MechJeb calculates the delta v for this rocket as if all the fuel will be burned in Stage 1 before the tanks are dropped, and therefore shows no delta v for Stage 0.

http://i.imgur.com/gb11Vmb.jpg

Thank you to all of the MechJeb developers! You guys are incredible.

Same issue I have, however it also have issues with radially connected SRB, stacks with SRB works well.

And yes keep up the good work

[Galane]

Same ship as before, http://pastebin.com/Q0BC0rtM with build 438. Taking note that the default ascent profile "Saner default ascent profile for 1.0" has been changed, I deleted the auto generated MechJeb cfg files. Odd thing, the turn start altitude, shape etc are all exactly the same as with build 435.

That rocket launches exactly the same as with build 435. TV limit doesn't appear to work and it gets going so fast that through the gravity turn it's going up sideways for a bit, like skidding around a turn. Perhaps some tail fins might keep the turn under control. Corrective steering didn't seem to make a difference.

That log thing. http://www.partsbyemc.com/output_log.zip

Landings on Kerbin? Still a no-go. Starts way too late, does a massive burn, undeshoots the target then flops back and forth doing "corrections". Sometimes it gets the blue mark near the red target but by then it's overflying the target.

How about simplifying the problem? Instead of attempting to account for all the new aero effects with varying angles and how parts now shield one another from drag, project a perfect, tail first landing trajectory from the space/atmo intersect. Next step, use all the control authority the craft has to attempt to hold the attitude to stay on that trajectory. If it yaws off due to aerodynamic force, (bleep)it, it's gonna crash same as a real one would in that situation. Needs moar reaction wheels and RCS!

Simplify the drag modeling MechJeb does. Start from a plane at the bottom of the craft, perpendicular to its axis, and project upwards to calculate the cross sectional area. Do math with that and the air density gradient. You're looking for optimal. If the rocket is built to where it can stay at optimal orientation with MechJeb at the controls, it should land intact and close to target.

If the rocket doesn't have enough control power to keep itself at optimal orientation, the landing will be sub optimal or *splat*, same as a real tail first landing rocket would be.

It doesn't have to be MagicJeb, able to save any craft that hasn't broken apart from tumbling out of control.

Another way to look at the problem. The design of bullets. People who design the shapes of bullets don't care what happens aerodynamically after one tumbles. If it tumbles, the shape and/or mass distribution was sub-optimal. Their problem is designing bullet shapes to keep them from tumbling.

KSP's new aero model is just going to force people to design more stable rockets for landing on bodies with atmosphere. MechJeb doesn't need to be able to safely bring down unstable rockets. "Aerodynamic forces exceeded available control authority. Rocket tumbled out of control and crashed."

[Aspira]

You usually have to tweak the ascent curve to fit your rocket. Some rockets require a smoother ascent profile, and that seems to be the case with yours.

How do I do that. Tryed every possible combination in the edit ascent path window and no matter what it flips 6-7km after launch.

[futrtrubl]

Terminal Velocity limit I think does still work, it's just that terminal velocity is much much higher than before if you have an actually streamlined craft.

[sarbian]

Exactly!!! If the landing autopilot can't even put you on the correct course without burning every last drop of fuel in your tanks AND still leave you with a 20km periapse on the opposite side of the planet from your target then there is clearly an issue!!!!!

Ok, let me deal with that first. Each time I see an exclamation point in that thread I want to go to my couch, launch my console and finish a JRPG or bindge on Netflix. For those who are not familiar with JRPG they takes a good hundred hours to finish. Time that will not be spent on MJ.

I get that some of you are frustrated but if you must vent don't do it here. I have been working on those update non stop for nearly 3 weeks now so the last thing I need is people who just come to complain that a major rewrite of the code is not perfect on the first release.

Landing Guidance -> Land At Target still appears to be broken with build #438. This is with Vanilla install, sandbox game. Again, I have no idea where any logs you'd like to view are located. Sorry. I'm not deliberately trying to be ignorant; I simply am. =)

The requested file lives here https://www.dropbox.com/s/hhozdbr0rx7qsx9/output_log_landing.txt?dl=0

I will have a look tomorrow.

I believe this is a bug:

The Delta-V Stats for designs with drop tanks and fuel lines don't correctly separate the stats for before and after the tanks are empty. Example: MechJeb calculates the delta v for this rocket as if all the fuel will be burned in Stage 1 before the tanks are dropped, and therefore shows no delta v for Stage 0.

Yes, the KER code I now use use a simpler logic for staging. Bringing back the old MJ staging locgic is on my list of short term fix

Hi everyone, I've only been playing a short time but I'm trying to get mechjeb to work. I'm on career mode. I have it set to autostage, avoid overheating, turn at 11k, and orbit at 100k. It takes off fine, it gets to suborbit fine, then it starts it's circularization. However, halfway into circularization there is a stage change and as soon as the second rocket is set to kick in, the autopilot turns off . I've tried everything I can think of. I could try to redesign the rocket to get rid of that stage change, but I'd rather just solve the problem.

Like I said, I'm new here, so if this is the wrong forum let me know.

Thanks!

The best was for me to help would be for you to post your ship/save somewhere so I can duplicate your problem.

Aspira : make sure the limit AoA is active it should help with most case. As the other said the problem is that the rocket turn too hard

Galane : the problem is already simplified. Have a look at the code before telling me how to change it...

[Aspira]

Aspira : make sure the limit AoA is active it should help with most case. As the other said the problem is that the rocket turn too hard

AoA is active. Tryed every angle 1-20 and it still flips 6-7km after launch. What bugs me is that in version 435 it was working for me after I find that TV was not working (posted at steam forum). Someone point me here and find out about 438, after that I cant get it to work. Clean install and now 435 wont work either. The problem for me dont seem to be AoA its not loosing control when it turns. Its loosing control while on vertical ascent. When prograde starts to move mechjeb is not following it.

Love Jrpg.

Thanks for your reply and effort just after 1.0 release.

Care not about imediatist (not sure if this is a word, but I am pretty sure you got it)

[edemlama]

Terminal Velocity limit I think does still work, it's just that terminal velocity is much much higher than before if you have an actually streamlined craft.

no it doesn't, limit to terminal velocity has a horrible throttle control and it is basically unusable

[Galane]

Tonight I had it almost nearly doing successful landings using land somewhere instead of trying to land at a target. The issue is having enough fuel and thrust to slow down enough to avoid too fast of an impact. May have to resort to using parachutes...

The lander atop these two rockets can *almost* land from a 71KM Kerbin orbit using Land Somewhere. That just aligns for the deorbit burn and burns. No turning back and forth, no corrections, no wild things like trying to reverse to a retrograde orbit or drastic changes in inclination.

http://partsbyemc.com/pub/two-step-lander.zip

They're a pair of old rockets, slightly modified to make them launchable in KSP 1.0. An all rocket SSTO and the same design modified with staging. I did the SSTO just to see if I could do it. That lander design (with vastly different boosters and transfer vehicles) has been to Mun, Minmus, Duna and Ike. They have a low enough TWR that terminal velocity and heating while launching isn't a problem.

[BudgetHedgehog]

Thanks for updating, sarbian - having used FAR before, I completely ignored MJ in atmo anyway, but in space, it removes a lot of the tedium. Many thanks again