Jump to content

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


r4m0n

Recommended Posts

Thanks, It bugs nicely here too :)

I'll look into it.

I did some more testing and found that limiting the model to a single parachute avoided the Hyperbolic orbit for the predicted landing point. The max drag Gs was still corrupt. Turning off Deploy Parachutes corrected both issues.

I hope that this information is helpful to you.

skips

Link to comment
Share on other sites

OK, some weirdness with the latest dev build... when I select Landing Gudance/TargetKSC/Land at Target rather than "predicted landing error", I get a long-a-- number (five-six lines) tagged "hyperbolic error", and it just burns and burns and burns... I had to abort the autoland, and when I selected land anywhere, I got the same thing. Eventually I had to just let the craft fall to 10km or so and then enable land anywhere.

Link to comment
Share on other sites

OK, some weirdness with the latest dev build... when I select Landing Gudance/TargetKSC/Land at Target rather than "predicted landing error", I get a long-a-- number (five-six lines) tagged "hyperbolic error", ...

Turn off Deploy Parachutes option in Landing Guidance.

Link to comment
Share on other sites

I did some more testing and found that limiting the model to a single parachute avoided the Hyperbolic orbit for the predicted landing point. The max drag Gs was still corrupt. Turning off Deploy Parachutes corrected both issues.

I hope that this information is helpful to you.

skips

Yes, it is really useful. It means I have to add ASAP the parachute fix that codepoet sent me :)

Link to comment
Share on other sites

Yes, it is really useful. It means I have to add ASAP the parachute fix that codepoet sent me :)

I was just about to say - that is caused by a huge deceleration when the parachute deploys (in the targeting simulation) being greater that the velocity, and so boucing the craft back out into space on a hyperbolic escape trajectory.

While you are waiting for a fix, you *might* be able to get around it by adding a drogue parachute to your craft to lessen the deceleration when the chute(s) deploy.

Link to comment
Share on other sites

I was just about to say - that is caused by a huge deceleration when the parachute deploys (in the targeting simulation) being greater that the velocity, and so boucing the craft back out into space on a hyperbolic escape trajectory.

I am inclined to disbelieve this explanation. The best that the braking drag could do is to stop the motion of the vehicle, which results in an elliptical orbit with a semi-major axis of zero. It would appear that the simulation is broken in this corner case. Further as it currently works the vehicle continues to attempt to decelerate. If it had "bounced" I would have expected the vehicle to flip over and accelerate.

skips

Link to comment
Share on other sites

Since the landing autopilot seems to be being actively worked on at the moment, would it be possible to get a version that doesn't land so much as crash?

I'm thinking for missile guidance mainly :)

Link to comment
Share on other sites

@NathanKell Yes it did show up. However there are a few pull request that affect a thrust/staging that I'll keep for after r4m0n release a news stable because they are impact more things in MJ.

@skips codepoet is right. The simulation recorded a force of nearly 300g when the chute opened. This made strange things append with the limited simulation precision ( it's the simulation run by MJ, not by the game ).

I just pushed the fix for parachute. The build bot will spew a new build soon.

In the end I simulate the parachute opening better and Codepoet fix is less needed (but still here because it's a good idea). With this fix the predicted max drag gees is 67 while the real drag gees was 64. Before I had a predicted max drag gees of 300+

@a__gun I won't for two reason : 1) breaking and running at full speed are two really different code. 2) I don't do weapon stuff.

Edited by sarbian
Link to comment
Share on other sites

Since the landing autopilot seems to be being actively worked on at the moment, would it be possible to get a version that doesn't land so much as crash?

I'm thinking for missile guidance mainly :)

Search the forum for KETM. I've been too busy with Kethane extraction to pursue orbit to surface high yield energy transfer utility. :wink: The S-KETM V4 does a fair job of decommissioning medium size and larger space vehicles.

Link to comment
Share on other sites

@sarbian Fine. I was going to suggest that there were couple approaches that you could use rather than attempting to simulate the parachute opening.

1. ignore the issue and treat the case the same way that you would handle the user activating the parachute deployment. I suspect the inaccuracies would not make a significant difference in the landing point.

2. move the simulated landing point up to the altitude that the parachute is planned for opening and change the destination velocity to the terminal velocity of the parachute deployed configuration.

3. move the simulated landing point to the altitude at which the parachute is planned for opening and use the specified destination velocity.

approaches 2 and 3 have the nice side effect that the G force on the vehicle is significantly reduced and the survivability of fragile vehicles would be significantly improved.

Just some thoughts,

skips

Link to comment
Share on other sites

@a__gun I won't for two reason : 1) breaking and running at full speed are two really different code. 2) I don't do weapon stuff.

1) I would of thought its a simple matter of dropping the 'turn retrograde and deccelerate' step..?

2) Interestingly the last time I wanted to make use of this feature was to lanch a orbit to ground supply 'missile' filled with KAS goodies so not weaponised at all :P

Search the forum for KETM. I've been too busy with Kethane extraction to pursue orbit to surface high yield energy transfer utility. :wink: The S-KETM V4 does a fair job of decommissioning medium size and larger space vehicles.

I found the relevant thread here. Looks interesting I'll give it a proper read later

Link to comment
Share on other sites

@skips most of those simplification works would work ok for Kerbin or Duna, but as soon as you go to Eve you need a better sim to get any precision. The denser atmo change a lot of things.

But the point is moot for now, my news code seems to work fine :)

Link to comment
Share on other sites

@skips most of those simplification works would work ok for Kerbin or Duna, but as soon as you go to Eve you need a better sim to get any precision. The denser atmo change a lot of things.

But the point is moot for now, my news code seems to work fine :)

Hopefully the new version works on Tylo too... MJ seems to have fits there.

Link to comment
Share on other sites

@skips codepoet is right. The simulation recorded a force of nearly 300g when the chute opened. This made strange things append with the limited simulation precision ( it's the simulation run by MJ, not by the game ).

I just pushed the fix for parachute. The build bot will spew a new build soon.

In the end I simulate the parachute opening better and Codepoet fix is less needed (but still here because it's a good idea). With this fix the predicted max drag gees is 67 while the real drag gees was 64. Before I had a predicted max drag gees of 300+

I just read your code - I am really impressed! I knew that my implementation had an error creaping in somewhere, but I could not work out where it was - but you had spotted it - it is the need to take account of the deployment speed which I had not even noticed in the object model.

Link to comment
Share on other sites

MechJeb 2.1 finally released :-)

Most interesting changes:

  • Full support for Career mode, with modules individually unlocking in several nodes;
  • New MJ2 Pod part;
  • Several bugfixes and improvements;

Link to comment
Share on other sites

MechJeb 2.1 finally released :-)

Most interesting changes:

  • Full support for Career mode, with modules individually unlocking in several nodes;
  • New MJ2 Pod part;
  • Several bugfixes and improvements;

WOOOOHOOOOOO!!!!

I was actually doing pretty well with career mode and no mechjeb for a while, but the repeated stuff and lack of precision when going between planets was getting annoying.

Link to comment
Share on other sites

1) I would of thought its a simple matter of dropping the 'turn retrograde and deccelerate' step..?

2) Interestingly the last time I wanted to make use of this feature was to lanch a orbit to ground supply 'missile' filled with KAS goodies so not weaponised at all :P

I found the relevant thread here. Looks interesting I'll give it a proper read later

The Armed Guard system can also be quickly adapted to a single purpose satellite launch. Remove the I-beam and reaction wheel then add two more batteries and a Kethane scanner or other sensor. That was how I launched my first Kethane scan of Kerbin... swords into plowshares etc. The AG rocket stack has also worked well for a wide variety of small payloads to LKO.

Is MechJeb 2.1 fixed so it won't fire engines higher up the stack - if they don't have a decoupler or separator in their stage - when it calculates it needs more thrust? In other words, will it never light up more than one stage at a time?

Can the DLL just be dropped in place of the previous one and still work with integrated MJ mods to control pods?

Edited by Galane
Link to comment
Share on other sites

I dont understand how mechjeb works in career mode I have the ar202 case unlocked but I dont get any of the normal menu's only the basic VAB stuff where is the ascent guidance and other tabs

Me too, I have played most of the way through career mode without mods just for fun, but now I want my mechjeb back. All I'm getting is data, no autopilot functionality at all. Frustrating.

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...