Jump to content

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

r4m0n

By r4m0n
in KSP1 Mod Releases

 Share

Recommended Posts

BARCLONE

BARCLONE

Posted
Posted

Second attempt:

Turn start = 0.5 Km (500 m)

Turn end = 125 Km

Turn shape = 75%

Acc limit = 13

Results: Turn did not begin (visibly) until about 2000 m, at a V of 50 m/s. MJ immediately began a reversal that started flipping around 2600 m, at a V of about 100 m/s. Rocket was still pointing upward at 3850 m, but V had already started dropping when I disengaged.

Link to comment
Share on other sites
Andon

Andon

Posted
Posted

The way I run MJ/FAR ascents is probably not the most effective, but it works.

Turn Start: 5km

Turn End: 75km

Turn Shape: 125%

You've got to type in the 125% yourself, but MJ's perfectly happy to use it once you do.

Link to comment
Share on other sites
rosenkranz

rosenkranz

Posted
Posted

With MJ/FAR I ussually do this

Turn start: 100m

Turn end: 60km

Turn shape: 70%

TWR limited to 1.6 at launch with acceleration clamped at 22 m/s2

This leaves the payload only needing about 100m/s dV to finish orbit insertion.

Andon? how do you get 125%? I didn't think you could go over 100%

Link to comment
Share on other sites
Shad0wCatcher

Shad0wCatcher

Posted
Posted

Simple Fix, rocket needs control surfaces near the bottom. What's occurring is drag is overcoming forward thrust / momentum and pushing your rocket out of control. Either a 3 or 4-way symmetry winglet setup will stabilize that. I found some massive drag issues with those conic adapters from KW rocketry and FAR; basically working as intended as air is slowing down across those surfaces and causing more drag than just having a tank alone. Also Taller rockets are much better than fatter ones (less overall surface area for drag).

Link to comment
Share on other sites
BARCLONE

BARCLONE

Posted
Posted

Interesting observation:

I let one of these flights go through to burn-out, and BANG if the second stage didn't fly as straight as an arrow! Took the lander up from about 11 Km to burn-out at 68 Km, and the lander's engine circularized the flight at the desired 200 Km. It went bonkers just before final cut-off, though, due to torque.

Turn start: 150 m

Turn end: 70 Km

Turn shape: 75%

Acc limit: 13

Link to comment
Share on other sites
BARCLONE

BARCLONE

Posted
Posted
Simple Fix, rocket needs control surfaces near the bottom. What's occurring is drag is overcoming forward thrust / momentum and pushing your rocket out of control. Either a 3 or 4-way symmetry winglet setup will stabilize that. I found some massive drag issues with those conic adapters from KW rocketry and FAR; basically working as intended as air is slowing down across those surfaces and causing more drag than just having a tank alone. Also Taller rockets are much better than fatter ones (less overall surface area for drag).

Interesting. As mentioned, this design flew on rails using MJ dev 149...

I do model rockets, even designing them with RockSim, and understand the "taller is better" idea. Generally, you're right. As for fins, well, with gimballed engines, you shouldn't need them. Falcon 9 doesn't use them, Delta 4 doesn't use them, Atlas V and Ariane 5 don't use them either...

Now, it would be interesting to see where the CG-CP points line up. It's definitely acting like the CP is ahead of the CG...

Link to comment
Share on other sites
BARCLONE

BARCLONE

Posted
Posted
Turn off corrective steering and try again. You've still got a very top heavy rocket with one engine on the bottom while in atmosphere running FAR. That just seems like a bad idea in general.

Just tried this...

The rocket looked really good until it reached about 3800-4000 m, with a V of 120 m/s, then started reversing. Another note, although it generally held an eastward heading, it danced around before going off-course.

Link to comment
Share on other sites
Shad0wCatcher

Shad0wCatcher

Posted
Posted

Oh I agree wholeheartedly Barclone; but launch that rocket again and pay specific attention to the drag on those conic reducers (the ones from KW you're using to increase then reduce the rocket diameter) I ran into the exact same issue using those on one of my own rocket designs. Drag coefficient using them absolutely skyrockets no matter where or how they're positioned on the rocket. (I've seen upwards of 100-150 kN of drag using the 3.75-2.5 as a reducer off a 3.75 meter expanded fairing)

Link to comment
Share on other sites
BARCLONE

BARCLONE

Posted
Posted (edited)
Oh I agree wholeheartedly Barclone; but launch that rocket again and pay specific attention to the drag on those conic reducers (the ones from KW you're using to increase then reduce the rocket diameter) I ran into the exact same issue using those on one of my own rocket designs. Drag coefficient using them absolutely skyrockets no matter where or how they're positioned on the rocket. (I've seen upwards of 100-150 kN of drag using the 3.75-2.5 as a reducer off a 3.75 meter expanded fairing)

Made some changes:

Swapped out the lower tank/transition to the NovaPunch "HH" series. Got some extra fuel in the swap...

Swapped the upper transition to the Keramzit Procedural Fairings type. Very explosive separation...

The rocket locked on rails from start to finish, and the circularization burn only took 100 m/s of Dv. Usually it was taking between 300 and 400 m/s for the previous version.

It still goes bonkers at the end of the burn...

Turn start: 150 m

Turn end: 70 Km

Turn shape: 75%

Acc limit: 13 m/s/s

I'll play with this for a while to tweak it out. It would still be interesting to determine how those drag numbers are affecting MJ so badly, though.

Side note to Sarbian: I saw your post about mentioning the CM icon to the other thread, but wanted to mention here that I DL'd a fresh copy from your link this morning, and the duplication issue has disappeared. I must have had an older build that had issues.

UPDATE: Ignore that last part. The CM icon is still duplicating. I'll post in the CM thread shortly...

Edited by BARCLONE
Link to comment
Share on other sites
Galane

Galane

Posted
Posted (edited)

The basic thing about rocketry is the center of gravity (or mass) should be ahead of the center of pressure. The classic way to check COP is to cut out a cardboard silhouette and check its center of gravity. Compare that location to the location of the center of gravity of the actual rocket.

Built that way, if the rocket veers off course, there's more drag behind the COM which forces the tail back into line. If the COP is ahead of the COM, there's more drag on the front end and it blows around farther. Rockets with a big nose and a small tail need either fins or additional weight at the tail. That's why payload lifting model rockets with capsules larger than the body diameter have large fins.

Checking the COP in the VAB could be done with a simple parallel raycasting from one side to create a virtual silhouette object then calculating its COM to generate an indicator to render. Dunno if a plugin could do that but it's something that would be quite useful if built into a future release of KSP.

Might be better to raycast from two directions, 90 degrees apart, then average the COP positions or show a high and a low position, since making rockets that aren't symmetrical in all directions seems to be a popular thing.

Once out of an atmosphere, the COP doesn't matter because there's no gas to cause drag.

Edited by Galane
Link to comment
Share on other sites
skips

skips

Posted
Posted

Just to add some wood to the fire, here are some of my recent observations with MJ.

1.) Launch a two stage rocket into Kerban orbit with MJ controllers on both stages. Separate the stages. The Detla-V Stats for stage 1 look normal. Switch to the decoupled stage 2 and the Delta-V Stats are not reasonable. The Mass is zero and the TWRs are all NaN. The vehicle statistics display has a reasonable mass and MJ can control the stage. (even to the point of successfully landing at KSC). The log at that point is a bit screwy but does not appear to be related.


[LOG 17:48:31.747] activating stage 1 - current stage: 2
[LOG 17:48:31.747] [science.module]: Activated
[LOG 17:48:31.748] Should have Jettisoned
[LOG 17:48:31.748] [liquidEngine3]: Activated
[ERR 17:48:31.780] MechJeb caught exception in core OnSave: System.NullReferenceException: Object reference not set to an instance of an object
  at MuMech.MechJebCore.OnSave (.ConfigNode sfsNode) [0x00000] in <filename unknown>:0 
[LOG 17:48:31.786] [00:06:26]: Separation of stage 2 confirmed
[LOG 17:48:31.786] [stackDecoupler]: Activated
[LOG 17:48:31.786] [mumech.MJ2.AR202]: Activated
[LOG 17:48:31.786] [batteryPack]: Activated
[LOG 17:48:31.786] [batteryPack]: Activated
[LOG 17:48:31.786] [batteryPack]: Activated
[LOG 17:48:31.786] [GooExperiment]: Activated
[LOG 17:48:31.786] [GooExperiment]: Activated
[ERR 17:48:31.836] MechJeb module MechJebModuleThrustController threw an exception in Drive: System.NullReferenceException: Object reference not set to an instance of an object
  at MuMech.CelestialBodyExtensions.RealMaxAtmosphereAltitude (.CelestialBody body) [0x00000] in <filename unknown>:0 
  at MuMech.VesselState.TerminalVelocityStockKSP () [0x00000] in <filename unknown>:0 
  at MuMech.VesselState.TerminalVelocity () [0x00000] in <filename unknown>:0 
  at MuMech.MechJebModuleThrustController.TerminalVelocityThrottle () [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 
[LOG 17:48:37.446] [FLIGHT GLOBALS]: Switching To Vessel Mun Surveyer 1 Probe ---------------------- 
[LOG 17:48:37.448] stage manager resuming...
[LOG 17:48:37.501] Focus changed! Forcing Mun Surveyer 1 to save

2.) Using the Return from Moon option of the Maneuver Planner to return from Mun appears to generate erroneous maneuver nodes. Starting from a 100 km equatorial orbit around Mun and requesting a 100 km periapsis gives a maneuver that results in an orbit that intersects Kerban. Reduce the Prograde velocity change by about 6 m/sec and the orbit grazes the Kerban surface. Another 6 m/sec and the periapsis is approximately 100 km. If you request a 300 km periapsis, the maneuver planner gives you approximately a 100 km periapsis. Repeat the experiment from Minmus and a 100 km request yields an orbit with a 100 km periapsis and a 300 km request yields a 300 km periapsis. It appears that the initial conditions for the Mun computation must be incorrect.

3.) When making inclination changes with the maneuver planner, you can match the orbital plane of another object orbiting the same central object or at some angle to the equatorial plane. There does not appear to be an option to set an inclination with respect to the central object's orbital plane. This option is equivalent to the equatorial plane, if the rotational axis of the central object is normal to its orbital plane, but this situation is rather rare in the real world. (Earth's orbital plane i.e. the ecliptic is tipped 23 degrees from the equatorial plane.) Matching the inclination of the orbit of the central body is useful for making maneuvers that are designed to move you out of that object's sphere of influence.

skips

Link to comment
Share on other sites
Virtualgenius

Virtualgenius

Posted
Posted (edited)

I will broach Sarbians favorite subject the landing AP it maintains roughly a 10m-50m difference from the target right up until the last stages before landing it then blows the landing out by 200m. So i adjusted to allow for that and it still did it, do you think it miscalculates its speed and trajectory in the last stages of landing it seems to start with a small aerobraking burst then settles into a 200m miss before landing perfectly. This is on kerbin using a home module at the north pole dll version 150 start orbit 300km

Edited by Virtualgenius
Link to comment
Share on other sites
DomDiaemus

DomDiaemus

Posted
Posted

Rover module seem a bit useless... it can't deal with tiny elevations (by going around in circles) and often with rover at max speed which obviously is a recipe for disaster.

Is there a way to control the rover max speed???? So not to go to max but be set at max specified by the player? Thanks

Link to comment
Share on other sites
BARCLONE

BARCLONE

Posted
Posted

I tried building a "long and narrow" vehicle with the same fuel capacity and the same engine, using KW's 2.5m parts. Before it reaches 5 Km, it starts dancing around prograde, but somehow managed to reach 70 Km and deliver the second stage to orbit. It never stabilized until after staging...

I tried adding some torque to the first stage with an SAS module (spinning wheel thingy). Between 5Km and 10Km, it lost it and flipped. Even started pitching toward 0 degrees (due north) before flipping...

I tried adding 4 fins in addition to the SAS. It still danced and flipped...

I tried adding another SAS to the booster, plus the fins. It still danced and flipped...

The problem seems to get progressively worse with each attempt...

Watching the Pitch/Roll/Yaw needles, it appears MJ is not letting up on the control inputs until well after the rocket has passed the "neutral" zone. It is still trying to correct for an error that has already been corrected, thus over-correcting and inducing the flip. Can the amount of input be toned down? Or can the inputs be neutralized quicker? Is this a "tweakable" value?

Link to comment
Share on other sites
Guest

Guest

Posted
Posted (edited)
I tried building a "long and narrow" vehicle with the same fuel capacity and the same engine, using KW's 2.5m parts. Before it reaches 5 Km, it starts dancing around prograde, but somehow managed to reach 70 Km and deliver the second stage to orbit. It never stabilized until after staging...

I tried adding some torque to the first stage with an SAS module (spinning wheel thingy). Between 5Km and 10Km, it lost it and flipped. Even started pitching toward 0 degrees (due north) before flipping...

I tried adding 4 fins in addition to the SAS. It still danced and flipped...

I tried adding another SAS to the booster, plus the fins. It still danced and flipped...

The problem seems to get progressively worse with each attempt...

Watching the Pitch/Roll/Yaw needles, it appears MJ is not letting up on the control inputs until well after the rocket has passed the "neutral" zone. It is still trying to correct for an error that has already been corrected, thus over-correcting and inducing the flip. Can the amount of input be toned down? Or can the inputs be neutralized quicker? Is this a "tweakable" value?

Whatchoo tawkin' 'bout, Willis?

KW tanks and engines, 90km orbit with a little petrol in the tank to spare. First stage is a FASA tank, same capacity though and I like the markings. No flips, no shimmies, no problems. Oh! No fins, either. Probably one of the most inefficient launch vehicles I've built in a very long time. A bit of a nail biter on ascent as far as fuel consumption goes, but it worked just fine.

thdQ5rH.png

1WHZ5TA.png

j7iz1TA.png

kCQKK7p.png

OMfrhuU.png

Edited by Jack Wolfe
Link to comment
Share on other sites
BARCLONE

BARCLONE

Posted
Posted

I've got an obwious scwewy wabbit installation, with mods acting ugly and not playing well together. I just added some of the biggest fins in the collection to the rocket, and it still flips before it reaches 3Km.

It's "zero room" time, Jack. I've got to strip it down to just the essentials again and see what happens. Glad to hear someone is getting their ships to work. Just wish it were me...

Link to comment
Share on other sites
Shad0wCatcher

Shad0wCatcher

Posted
Posted

Definitely sounds like something borked over there Bar :(. Mine have been doing some pirouetting depending on build; but I haven't built any decent rockets lately (doing too much editing of mods to make them fit better together), and I've just been launching essentially stub-rockets. They're ugly as sin but 3+ TWR and 5k+ DV just to get to orbit using FAR (READ: severe overkill) and test things.

Link to comment
Share on other sites
Galane

Galane

Posted
Posted

For my test crash last night. Ship consisted of this stack, bottom to top. Skipper, X200-8, 2-man lander can, big reaction wheel and shortened FL-R1 from ReStock, orange tank, 4K battery, large RGU, 3-in-1 Sr. sized docking port. Stuck to the sides were a pair of VTOL mounts from Infernal Robotics (attached with hydraulic detachment manifolds) with Cirrus envelopes, pivoted down. Four radial chutes at the top of the orange tank. Had a ring of 6 OX-Stat panels and 8 5-way RCS blocks from ReStock mounted here and there.

The landing gear was 10 short I-Beams hung from the bottom edge of the X200-8 with LT-2 legs.

Landing was proceeding fine, chutes out, slowing down nicely, just needed enough push from the engine, soon enough, to bring it down slow enough. MechJeb didn't drop the landing gear until the engine nozzle was about to touch the ground. The gear hit the ground before it fully extended.

Was it mis-computing some lift that wasn't there from the fully collapsed Cirrus gas bags? Why would it wait until too late to extend the landing gear? If it had extended the gear high enough the landing would've been hard but most likely nothing would have broken.

The crash lost two legs (but the I-beam connections held thanks to the latest Kerbal Joint Reinforcement) and the orange tank plus everything attached to it except the VTOL rotatrons and the Cirrus bags was destroyed.

I'm attempting to put together a minimum weight lander with gasbags to lift it to a height where it can make it back to orbit so I won't have to launch a crazy massive craft and transfer it to Eve. Problem is MechJeb keeps cutting the final burn too close unless the lander is way overpowered.

Link to comment
Share on other sites
Shad0wCatcher

Shad0wCatcher

Posted
Posted

I've noticed sort of the same thing Gal using just "land somewhere" that it does a suicide burn to kill velocity versus choosing a landing site where it'll actually perform a normal approach above then kill horizontal vel; then a gentle descent.

Link to comment
Share on other sites
Guest
This topic is now closed to further replies.
 Share
Go to topic listing
×
×
  • Create New...