[WIP] Infernal Robotics - Next

[Rudolf Meier]

now I like the solution... at least in first tests... unfortunately the linear limits are completely wrong calculated now and I've to fix that first... but that's just setting the anchor and connectedAnchor correctly... I did mess up somethign with those values

and another thing that's not good... when you have too much load on a joint, it bends (that's pretty good how it does it) but when you load such a scene... the parts start move around the body... that's fixed for rotational joints, but not for translational joints, since it was never ment that they could "rotate"... but, in case they bend heavily... that's something I need to solve too...

[Rudolf Meier]

the question here is: is this an IR problem? ... I guess not... those part-attachment-point-drifts I see all the time with all sorts of parts (not so much for stock parts) ... but in case something "bends" (happens very rarely for stock parts as I found out), then... they jump back into their position without force applied when you start time warp or save/load ...

I don't like it, but... it's the game... maybe our parts should act in the same way... that should be possible to do... I don't know how they do it, but I will find out (launch clamps for example do this) ... the rest I think is up to KSP developers to fix that one day...

[Rudolf Meier]

in a way it's unbelievable... why did never ever someone of the development team think about the fact, that somewant wants to save/load/timewarp a ship that has a force applied to a joint and that this force is bending this joint? ... that should be something that's inside the game already... that's basic stuff

...  but as I said ... I'm used to things like that. Today (in my job) we had to store a "string" in a "boolean"  ... otherwise the compiler would have selected a wrong option... took us some hours to find that out

 

Edited March 20, 2018 by Rudolf Meier

[Rudolf Meier]

ok, here is the basic problem behind all this

do the following experiments

take a launch clamp, attach an I-beam or other truss (stock) to it so that it's pointing outwards (maybe you can add 2 of them) and then you add a heavy tank (or more) ... what you see is, that the launch clamp is bending and the tank will point a little bit down after a while

then do this with Extendatrons attached to a solid tower (not a launch clamp)... you will also see that it will bend and the tank points to the ground after a while

and after that do this with old Extendatrons too

... if you do now go to time-warp or do a save/load, we have different behaviour. The stock part will jump back and fall down again, the new Extendatrons will bend further and further and the old Extandatrons shift their attachment points towards the ground more and more with every step

 

And now we need solutions... jump back is the default behaviour (that's why they added the "physics easing" I think, because ships would break after loading otherwise) ... but I'm not sure if that's the way to go... like this every robotic arm/joint/part will jump into the "non-bended" position as soon as you do time-warp or something like that... and this would give us very strange results I think... loading a scene with a part attached to an IR part would almost immediately destroy everything.

If the joint is bended along the axis it is moving... then that's no problem. This should work in IR Next already. But if it's bended to the side... well... then I do have a little problem... because to fix this I would need to put some force on the joint... and doing this means that I need to let it rotate on those axis... but then... I'm not sure if I can still have a locked setting on that axis... *hmm* you know what??... I should try something  ...

thanks... it always helps to discuss such problems...

[kcs123]

Uh, I need one more coffee to be able to diggest everything said

But, it is always nice to share thoughts, sometimes some crazy nonsense ideas may be inspiration for better solution, even when such idea is no plausible at all.

[Rudolf Meier]

4 minutes ago, kcs123 said:

Uh, I need one more coffee to be able to diggest everything said

But, it is always nice to share thoughts, sometimes some crazy nonsense ideas may be inspiration for better solution, even when such idea is no plausible at all.

That's true! Sometimes when we don't have useful ideas for problems in software development, we explicitly start implementing "stupid" or "can never work" ideas... in almost every case we found a good solution for the problem after just one or two hours of development... but, try to do this in "normal" companies... impossible

[Rudolf Meier]

5 hours ago, Rudolf Meier said:

*hmm* you know what??... I should try something  ...

so far I like what I found...  ... I will build something new now...

... *hmm* not every idea works. I need more time to find out if there is a solution that provides everything we need... which is not too strong servos but still not shifting in non-movable directions

the point is, one solution seems to be ok for me... but the non-moving mesh is slightly out of position... I could try to adjust this

but that's not all... there is also the problem that parts attached to IR parts seem to have a problem with the bending of the IR parts... so, maybe it would be better not to allow them to bend... the problem here is: how do we do this? it's not working without making the joints infinitly strong... and then there is never something like a "stuck" servo ... it would always have infinite torque

... but, maybe I'm wrong and there is a setting that allows both things. I'm still searching...

Edited March 20, 2018 by Rudolf Meier

[Rudolf Meier]

it is not easy, but, the old IR is doing some kind of "active correction" of the wrong positions... that's something I did ignore completely... what it does is, allow movements along all axis and then point into one direction and set the target position in the space and not along the axis joint... that's probably the main difference and why it seems to be that stiff

one of the questions here could be... if we like that and what the difference between this and an ibeam is... because the ibeam seems not to bend that way and does not have an active compensation of the wrong position... so... lots of data and cases to compare

[kcs123]

More and more you dig, you will most probably found even more hacks that someone put there to workaround of unity problems. Some made by Squad, some by other moders.

[Rudolf Meier]

I need to restart my investigation of translational joints... I have no idea why joints, locked on y, z axis and with rotation on all axis locked do rotate and move along those axis, even when they are infinite strong.. that doesn't make sense to me. ... so, no solution up to now...

the old IR joints are extremely strong... but I haven't found out why... they don't move a bit into the wrong direction...  that's incredible... every joint (every instance) has it's axis pointing to somewhere else... I don't see a system here... it's almost random. But, they keep their translational positions away from the movement axis 100% ... they don't move a bit... I worked with IR for a long time now... but I still don't see how this is done... must be a stupid detail I'm missing

Edit: I found one detail... I'm trying this tomorrow... and I'm trying if... KJR (and therefor also IR) could be improved in 1.4.1, when I'm using a new setting... it's still a mistery why some joints are stronger than others... it is clear, that it has to do with weight... but maybe not with the absolute mass, but just with relative weights? ... I'm not sure, but I do have the impression, that this theory could at least contain some truth...  we will see...

Edited March 22, 2018 by Rudolf Meier

[edemlama]

dunno if it happened to anyone else but the latest legacy version broke my saves

[Rudolf Meier]

4 hours ago, edemlama said:

dunno if it happened to anyone else but the latest legacy version broke my saves

IR 2.x ?

[Rudolf Meier]

one question... if I could make the joints stronger either with performance penalties or only in 1.4.1 ... what do you prefer? drop support for 1.3.x and before or add the system with performance penalties? (I don't have the solution at the moment, but, as hypothetical question)

 

[ZodiusInfuser]

4 minutes ago, Rudolf Meier said:

one question... if I could make the joints stronger either with performance penalties or only in 1.4.1 ... what do you prefer? drop support for 1.3.x and before or add the system with performance penalties? (I don't have the solution at the moment, but, as hypothetical question)

 

My vote would be for 1.4.1 only, however I understand that people can still download the old version through the beta's tab in Steam.

[brusura]

11 minutes ago, Rudolf Meier said:

one question... if I could make the joints stronger either with performance penalties or only in 1.4.1 ... what do you prefer? drop support for 1.3.x and before or add the system with performance penalties? (I don't have the solution at the moment, but, as hypothetical question)

 

Always look forward....I would drop support for 1.3.x

[Rudolf Meier]

11 minutes ago, ZodiusInfuser said:

My vote would be for 1.4.1 only, however I understand that people can still download the old version through the beta's tab in Steam.

... I don't like dropping old versions too soon. I'm thinking about writing 2 versions. An optimized one for 1.4.1 and one for 1.3.1, 1.2.2 that does it with the performance penalties (which is the only solution for those versions) ... but...  ... don't know if that's a good idea

maybe I will write both and then compare them... if it's not too much, I could try to support the special one for 1.3.1 for a while...

but first I'm now verifying, if my theory is correct with 1.4.1

Edited March 22, 2018 by Rudolf Meier

[Rudolf Meier]

ok, I can confirm, that I can make the joints very very very strong now without adding mass somewhere... but, just in 1.4.1 and currently I don't understand why it works... I simply multiplied a value by 10... but I expected, that it should work without this multiplication... anyway, that's not so important, I can figure this out... at least we know, that it works somehow... if this will bring us a usable overall situation... no idea... but maybe... at least it's looking better than yesterday

now I will try to implement the 1.3.1 solution... if this can be done easily, then... I might support this too, otherwise... I will re-think the idea

[edemlama]

8 hours ago, Rudolf Meier said:

IR 2.x ?

I'm using the latest beta downloaded from this thread 

[Rudolf Meier]

2 minutes ago, edemlama said:

I'm using the latest beta downloaded from this thread 

*hmm* ... how did it break them?

[edemlama]

3 minutes ago, Rudolf Meier said:

*hmm* ... how did it break them?

once I launch the game it tells me that the parts are missing, all the vessel using them are gone, but if I look in the VAB all the parts are there

 

[Rudolf Meier]

1 minute ago, edemlama said:

once I launch the game it tells me that the parts are missing, all the vessel using them are gone, but if I look in the VAB all the parts are there

 

I think I did change some values like torque and things like that a week ago... but... I didn't expect that this breaks the part. Because... the name should still be the same... at least I think my saves still load... I will do some tests later... but I wasn't aware of such a problem with KSP... but maybe I simply modified more than I thought... I will have to investigate this ... can you keep some old and new save files, just in case, so that I could compare them, when I'm running out of ideas?

[edemlama]

1 minute ago, Rudolf Meier said:

I think I did change some values like torque and things like that a week ago... but... I didn't expect that this breaks the part. Because... the name should still be the same... at least I think my saves still load... I will do some tests later... but I wasn't aware of such a problem with KSP... but maybe I simply modified more than I thought... I will have to investigate this ... can you keep some old and new save files, just in case, so that I could compare them, when I'm running out of ideas?

yes sure I can give them to you right now if you want

 

[Rudolf Meier]

22 minutes ago, edemlama said:

yes sure I can give them to you right now if you want

 

sure, if you can upload them somewhere... I'm currently working on the joint-strength problem, but I will try to look into your problem after that

[edemlama]

1 hour ago, Rudolf Meier said:

sure, if you can upload them somewhere... I'm currently working on the joint-strength problem, but I will try to look into your problem after that

here you go https://mega.nz/#F!LhEx2RRI!fYdaqdgfuLWnYKNy21LvLQ

in the folder you will find the save files, the log and the IR folder from 1.3 and 1.4.1

[Rudolf Meier]

3 hours ago, Rudolf Meier said:

now I will try to implement the 1.3.1 solution... if this can be done easily, then... I might support this too, otherwise... I will re-think the idea

it pretty difficult to turn around everything... I don't know if I will continue with this idea

and because of all those positive statements about the idea of dropping support for old KSPs and focus on a good solution for 1.4.1 I think I will stop trying to fix this... soon