Jump to content

[old thread] Trajectories : atmospheric predictions


Youen

Recommended Posts

Is there any way to increase the precision of drag calculations during re entry. I keep having an issue where I line up the red X to the desired way-point however on re entry the deceleration keeps making the X drift and I end up missing the target. Even if the engine had not already been jettisoned, its not possible to reorient for a correction burn during reentry without destroying the ship. So how exactly do I compensate for this drift prior to reentry? Normally it doesn't matter because there is a significant margin of error around most way points however some contracts like the one I am currently doing requires the pod to splash down precisely in a tiny pond >.<

Edited by Mikeloeven
Link to comment
Share on other sites

Hello,

first of all thanks for this wonderful mod that has been serving me well for the past few years. Anyway now I am playing with KSP 1.2.2 and devel version of FAR and I am getting somewhat inconsistent results with the calculated trajectories (I end up missing the target, going back to orbit, etc ...). It seems to systematically overestimate drag. Am I the only one to be affected by this kind of issue?

Edited by LastStarDust
Link to comment
Share on other sites

4 hours ago, LastStarDust said:

Hello,

first of all thanks for this wonderful mod that has been serving me well for the past few years. Anyway now I am playing with KSP 1.2.2 and devel version of FAR and I am getting somewhat inconsistent results with the calculated trajectories (I end up missing the target, going back to orbit, etc ...). It seems to systematically overestimate drag. Am I the only one to be affected by this kind of issue?

Same with me, its become so inaccurate in 1.2.2 that I finally just stopped using it.  

Link to comment
Share on other sites

On 4.3.2017 at 9:36 PM, dlrk said:

Any news on that targeting bug or those features I suggested?

Sorry, which targeting bug do you mean? I'm afraid I didn't have time for implementation of those features, and I also can't give you an ETA. RL is in the way :wink:

 

On 4.3.2017 at 11:35 PM, Mikeloeven said:

Is there any way to increase the precision of drag calculations during re entry. I keep having an issue where I line up the red X to the desired way-point however on re entry the deceleration keeps making the X drift and I end up missing the target. Even if the engine had not already been jettisoned, its not possible to reorient for a correction burn during reentry without destroying the ship. So how exactly do I compensate for this drift prior to reentry? Normally it doesn't matter because there is a significant margin of error around most way points however some contracts like the one I am currently doing requires the pod to splash down precisely in a tiny pond >.<

On 10.3.2017 at 8:36 AM, LastStarDust said:

first of all thanks for this wonderful mod that has been serving me well for the past few years. Anyway now I am playing with KSP 1.2.2 and devel version of FAR and I am getting somewhat inconsistent results with the calculated trajectories (I end up missing the target, going back to orbit, etc ...). It seems to systematically overestimate drag. Am I the only one to be affected by this kind of issue?

Nope, you are not the only one. Accuracy is broken. There is a bug report: https://github.com/neuoy/KSPTrajectories/issues/84 which shows that even "crafts" consisting of a single Mk1-Pod don't land at the predicted position. I'm afraid I don't have the in-depth code knowledge required to fix these kinds of issues. That's something that only @Youen can do, so we'll have to wait for him.

I'd say the mod is still *somewhat* useful for getting a rough landing estimate - but definitely not as the spot on targeting system that it used to be.

Edited by Kobymaru
Link to comment
Share on other sites

it could be just a change is air densety
if that is hardwired in the mods formula

do you know inaf to fined wat number are hardwired and wat the mod gets from the game?

edit:

i had a look and i think ther are only 2 number not from the game in stok AerodynamicModel/StockAeroUtil
double liftQ = dyn_pressure * 1000; but i think its only for scale (K)

and
double dyn_pressure = 0.0005 * rho * v_wrld_vel.sqrMagnitude;

 

in Trajectory.cs
i found
TotalForce += bodySpaceVelocity * (dt * 0.000015); // numeric precision fix

well that all i can do
but the line in Trajectory.cs looks promising

Edited by danielboro
Link to comment
Share on other sites

I am not sure if this is related to the accuracy bugs but once i started going to mun or minmus I found a major bug where the mod will throw a hissy fit when trajectories become hyperbolic and have no end point Basically this causes a huge framerate loss and frequent client crashes when plotting hohmann or interplanetary courses It got to the point where I have to uninstall the mod any time I am flying a mission that leaves kerbin orbit and reinstalling the mod once i get into orbit around the target.

Link to comment
Share on other sites

5 hours ago, Kobymaru said:

That's something that only @Youen can do, so we'll have to wait for him.

That's a bit of an overstatement... Unless you mean my code is so awful that no one can understand it ;-)

Unfortunately, I've lost interest in KSP in general for quite some time now, and I don't really want to reinstall the latest version just to fix bugs and upgrade code for the latest version, which is precisely the part I dislike about modding.

Maybe I'll get back to it later, but no ETA nor promises.

In the meantime, I can at least update the opening post inform people of what works or not, and on which version of the game. I think the most problematic issue is that prediction is not accurate, but I don't know what is the last version of KSP for which it worked correctly?

Also, despite what @Kobymaru says, I'm sure there are a lot of people around here that are able to improve the mod (starting with himself). And I'd be happy to release the work of anyone who would like to fix the issues / improve the mod. I can also give direct access to spacedock and AVC, and I think moderators can transfer the opening post ownership, or failing that we could create a new thread, if someone wants to maintain the mod.

I'm really sorry to "let down" the current users of the mod, but fixing this kind of issue can be cumbersome, and I don't want to force myself to do it if I'm not even going to play the game after that.

Link to comment
Share on other sites

One minor thing I would like to see... would it be possible to allow the mod to generate the CCIP (Continuously Computed Impact Point, aka the red X on the surface of the planet in map mode that shows your actual landing point on the map, taking into account planetary rotation) even on planets without atmosphere? I mean yeah, the lack of atmosphere will make the trajectory the same as the stock rail projection, but even so the stock model doesn't take into account the rotation of the planet. It would make pinpoint landings on planets or moons without atmospheres much easier.

Also while on the subject of the CCIP I would also like to see the possibility of it being projected onto the terrain in Flight Mode, augmented-reality style, so that you can see the affects of minor adjustments to trajectory without flipping back and forth from Map and Flight modes. This, again, would benefit people shooting to land on a dime (i.e. landing directly at KSP's launch pad to maximize return on reusable stages, or within 50m of a planet or moon base to facilitate refueling and resource transfers).

Link to comment
Share on other sites

1 hour ago, Youen said:

That's a bit of an overstatement... Unless you mean my code is so awful that no one can understand it ;-)

I tried to dig in the drag-calculation code, but that led to some serious confuzzlement :confused: Maybe I need to take a harder look.

 

1 hour ago, Youen said:

Unfortunately, I've lost interest in KSP in general for quite some time now, and I don't really want to reinstall the latest version just to fix bugs and upgrade code for the latest version, which is precisely the part I dislike about modding.

That is sad to hear, but I understand you very well.

 

1 hour ago, Youen said:

In the meantime, I can at least update the opening post inform people of what works or not, and on which version of the game. I think the most problematic issue is that prediction is not accurate, but I don't know what is the last version of KSP for which it worked correctly?

Good question, I will try to look into that. I have a "library" of KSP versions :wink:

1 hour ago, Youen said:

Also, despite what @Kobymaru says, I'm sure there are a lot of people around here that are able to improve the mod (starting with himself). And I'd be happy to release the work of anyone who would like to fix the issues / improve the mod. I can also give direct access to spacedock and AVC, and I think moderators can transfer the opening post ownership, or failing that we could create a new thread, if someone wants to maintain the mod.

Let's not go the thread ownership route, since we did that already (back and forth :wink: ). Turns out that I am also somewhat prone to going MIA for some time periodically.

But since I at least play the game myself, I guess it makes more sense for me to take over the maintenance. I would like to make changes to spacedock and AVC and preferably make releases on GitHub too, so shared ownership (if that is possible) would be nice.

1 hour ago, Youen said:

I'm really sorry to "let down" the current users of the mod, but fixing this kind of issue can be cumbersome, and I don't want to force myself to do it if I'm not even going to play the game after that.

No worries. If it stops being fun, there's no point in doing that.

 

This is a callout to everyone:

Help is very much wanted. While I can not make any promises or ETA's on new features and bugfixes, I can make promises that I will try out and release the changes that are sent to me. So please, dear Trajectories-Users, especially the ones with Programming experience: if you want to keep this mod working, please take a look at the code and try to improve it :)

1 hour ago, Auriga_Nexus said:

One minor thing I would like to see...

No offense dude, but I think this is one of those times where people need to "read the room".

As you can see if you read the last posts, we are struggling to keep this mods basic functionality afloat. Piling on Feature requests does not particularly help in this situation, because our priority right now should be: make it actually do what it's supposed to at its core.

As I mentioned above, contributions are very welcome. So if you really need your CCIP functionality, you should consider implementing it yourself. I am very happy to merge good Pull Requests and create releases for new features.

Link to comment
Share on other sites

3 hours ago, Kobymaru said:

No offense dude, but I think this is one of those times where people need to "read the room".

As you can see if you read the last posts, we are struggling to keep this mods basic functionality afloat. Piling on Feature requests does not particularly help in this situation, because our priority right now should be: make it actually do what it's supposed to at its core.

As I mentioned above, contributions are very welcome. So if you really need your CCIP functionality, you should consider implementing it yourself. I am very happy to merge good Pull Requests and create releases for new features.

Offense taken, how was I supposed to know you were going to get butthurt over a minor suggestion. But fine. I'll drop it.

Link to comment
Share on other sites

12 hours ago, Kobymaru said:

Sorry, which targeting bug do you mean? I'm afraid I didn't have time for implementation of those features, and I also can't give you an ETA. RL is in the way :wink:

I was referring to the accuracy bug. Take your time, hopefully we can get it fixed when RL moves out of the way. I'd help, but I have very minimal programming experience, and would probably do more harm than good. One possibility I would raise though, is that IIRC, Trajectories uses the FAR plugin to simulate reentry and landing, so maybe it's an issue on FAR's end?

Edited by dlrk
Link to comment
Share on other sites

6 hours ago, Auriga_Nexus said:

butthurt over a minor suggestion

Like I said: suggestions are great, but if developer time is in short supply, one must prioritize.

 

5 hours ago, dlrk said:

Trajectories uses the FAR plugin to simulate reentry and landing, so maybe it's an issue on FAR's end?

I think there is *also* an issue with the FAR-integration, but that is independent from the stock issue. Stock Aero prediction only uses Stock code, not FAR.

Link to comment
Share on other sites

At the end, it is mod developer who is offering his free time for making a mod like this. 

So you can't really demand stuff. I'm grateful they even want to do things like this. It makes the community something special. Keep up the good work guys!

Link to comment
Share on other sites

If the accuracy with stock aero is independent from use with FAR, that would explain why I don't seem to be experiencing the accuracy bug as badly/at all some times. Also means that using FAR is a work around

Link to comment
Share on other sites

Hey guys, I need a little bit of help in debugging.

What was actually the last version of KSP and/or the last version of Trajectories, where the Impact prediction actually worked correctly?

How to test:

  1. Load up a legacy version of KSP, find the appropriate release of Trajectories here: https://github.com/neuoy/KSPTrajectories/releases
  2. If you don't have Hyperedit, take the Kerbal 1 (Stock Craft), remove the parachute. If you do have Hyperedit, create a new craft and add in just A Mk1 Command Pod
  3. To the previous craft, add
    1. A TR-2V Stack Decoupler, upside down
    2. An FL-R10 RCS Fuel Tank
    3. 4 RCS Blocks on the RCS Fuel Tank
  4. Launch or hyperedit the craft into a 85x85 km orbit
  5. Lower your periapsis to 55km using the RCS thrusters
  6. Decouple everything so that only the Capsule itself is left
  7. In Trajectories:
    1. Next to "Descent Profile", click Retro
    2. Open the Target Tab, click "Set current impact"
  8. Timewarp to atmosphere, make sure you are pointed retrograde. In some KSP versions this is rather annoying due to broken SAS. Keep it pointed retrograde
  9. Observe if the Predicted impact point (Red) deviates from the Target impact point, and if you "hit" the target impact point.

For our purposes, everything within 5km is called a hit.

I actually went back as far as KSP 1.0.5, and even there the deviation was up to 30 km - with a craft as simple as a single Mk1 Command Pod. So when was this Mod actually accurate?

Link to comment
Share on other sites

You have accuracy issues with KSP 1.0.5 and Trajectories v1.4.5 ?

That's weird, because the changelog says accuracy issues with stock aerodynamics were fixed in that release, so it should mean I had actually validated it was accurate after the fix.

Link to comment
Share on other sites

On 17.3.2017 at 9:03 PM, Youen said:

You have accuracy issues with KSP 1.0.5 and Trajectories v1.4.5 ?

That's weird, because the changelog says accuracy issues with stock aerodynamics were fixed in that release, so it should mean I had actually validated it was accurate after the fix.

Yep, with v1.4.5 and 1.4.6. The impact reticle starts drifting away from the target reticle almost as soon as I hit the atmosphere and I end up around 30 km away from the initially predicted target - and that's with only the Mk1-Command Pod (or the Mk3-Command Pod) perfectly retrograde, so there shouldn't be any funky wing/lift business.

I'll go back more in history in the next few days and see when it starts being precise. Because I definitely remember times when I landed on the spot - I just don't recall when that was.

Edited by Kobymaru
Link to comment
Share on other sites

1 hour ago, Kobymaru said:

Yep, with v1.4.5 and 1.4.6. The impact reticle starts drifting away from the target reticle almost as soon as I hit the atmosphere and I end up around 30 km away from the initially predicted target - and that's with only the Mk1-Command Pod (or the Mk3-Command Pod) perfectly retrograde, so there shouldn't be any funky wing/lift business.

I'll go back more in history in the next few days and see when it starts being precise. Because I definitely remember times when I landed on the spot - I just don't recall when that was.

I concur, I remember using it on previous versions and it being quite accurate.  

Link to comment
Share on other sites

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