Rover Wheel Alignment Bug (?)

[NEBx]

Hi all,

 

I am having an issue where if I rotate rover wheels anywhere from 45 degrees from default orientation or greater, they will not accelerate my rover.  The wheels will turn but not accelerate.

I've done extensive testing and found this to occur even with good ground clearance, anywhere from one to all axles powered and regardless of vehicle mass or root part orientation.

I can't see a reason why wheels wouldn't be performing normally, in terms of acceleration, in any orientation because well... its a wheel. Screenshots below showing default orientation, 45 degrees and 90 degrees.

                                                        DEFAULT ORIENTATION (WORKS)                                                                                       45 DEGREE ORIENTATION (DOESN'T WORK)                                                                           90 DEGREE ORIENTATION (DOESN'T WORK)

 

Thank you for your time,

[War Eagle 1]

39 minutes ago, NEBx said:

Hi all,

 

I am having an issue where if I rotate rover wheels anywhere from 45 degrees from default orientation or greater, they will not accelerate my rover.  The wheels will turn but not accelerate.

I've done extensive testing and found this to occur even with good ground clearance, anywhere from one to all axles powered and regardless of vehicle mass or root part orientation.

I can't see a reason why wheels wouldn't be performing normally, in terms of acceleration, in any orientation because well... its a wheel. Screenshots below showing default orientation, 45 degrees and 90 degrees.

                                                        DEFAULT ORIENTATION (WORKS)                                                                                       45 DEGREE ORIENTATION (DOESN'T WORK)                                                                           90 DEGREE ORIENTATION (DOESN'T WORK)

 

Thank you for your time,

Probably not the reply you were hoping for but from what I've tested it seems if wheels are clipped or rotated off center then they won't move at all.
For example my tank in 1.0.5 had the wheels slightly clipped. It used to get up to about 22 m/s
In 1.1 the same tank won't move an inch. However if i go and fix the wheels so they don't clip it will move but it consumes so much electric charge that it gets a max speed of 5-8 m/s

Right now the wheels are plagued with bugs. 
Unless a moderater says other wise my advice would be just to play around with other things like rockets or maybe planes until all this gets fixed. 

If you use mods Kerbal Foundaries is working on updating to 1.1 and their makers has stated they got their tank tracks working so when they release their 1.1 version it may make for a good temporary fix. That is assuming they beat Squad to fixing the problem

[Majorjim!]

They work like this but each set of wheels are rotating in different directions. You need to use the widgets to reverse the drive of one of the pairs of wheels.

[katateochi]

1 hour ago, War Eagle 1 said:

Probably not the reply you were hoping for but from what I've tested it seems if wheels are clipped or rotated off center then they won't move at all.

Indeed, although these are two separate issues.  The clipped wheels not working (and being sunk into the ground) problem is hopefully something that will get fixed when they update to a new version of unity.
Quote from @Arsonide from this thread about the wheel clipping - http://bugs.kerbalspaceprogram.com/issues/8923

Quote

This is not a bug, and is quite intentional. We should be able to fix it in the next patch, when we upgrade Unity and can get proper clipping with the wheels.

 

The other problem that @NEBx is asking about (wheels rotated > 45 from ground) is I think something that we have to live with. It's a limitation with how wheels work (spoiler; they're not really wheels). see these two on the topic;
http://bugs.kerbalspaceprogram.com/issues/8344 
http://bugs.kerbalspaceprogram.com/issues/8172

[War Eagle 1]

40 minutes ago, katateochi said:

Indeed, although these are two separate issues.  The clipped wheels not working (and being sunk into the ground) problem is hopefully something that will get fixed when they update to a new version of unity.
Quote from @Arsonide from this thread about the wheel clipping - http://bugs.kerbalspaceprogram.com/issues/8923

 

The other problem that @NEBx is asking about (wheels rotated > 45 from ground) is I think something that we have to live with. It's a limitation with how wheels work (spoiler; they're not really wheels). see these two on the topic;
http://bugs.kerbalspaceprogram.com/issues/8344 
http://bugs.kerbalspaceprogram.com/issues/8172

So rather than putting off 1.1 until that issue could be fixed they decided to force a limitation and work on it later.
Im normaly not one to bash Squad but….thats a poor choice IMO. Better to push back a product that works than push one that will break things

[Majorjim!]

3 hours ago, War Eagle 1 said:

So rather than putting off 1.1 until that issue could be fixed they decided to force a limitation and work on it later.
Im normaly not one to bash Squad but….thats a poor choice IMO. Better to push back a product that works than push one that will break things

If the wheels are less than around 45 degrees they are spinning in different directions. Let me knowing reversing the drive of one pair of wheels helps. It should. 

[swjr-swis]

14 minutes ago, Majorjim said:

If the wheels are less than around 45 degrees they are spinning in different directions. Let me knowing reversing the drive of one pair of wheels helps. It should.

It won't matter. The problem is that code-wise, 'wheels' aren't actually round in the game, despite what your eyes tell you.

Ignore the round spinning visible part of the wheel that fools you into certain expectations, that is purely cosmetic. The actual wheel is an invisible skeleton without spinning parts so to speak that detects a nearby surface and creates a magical force under certain conditions. The code requires that the angle is close to perpendicular to the surface, or the 'wheel' is considered to not be touching and thus no force can be generated... despite the visual wheel spinning and touching or even clipping the surface and for all intents and purposes looking like it should be carving a ditch if it doesn't start moving soon.

 

Now cue the actual coders coming in to bap me upside the head and reassure everyone that Santa does actually exist, and wheels are in fact, really round and do spin, and we should all go back to sleep and all will be alright with v1.2 in the morning...
 

[Majorjim!]

Just now, swjr-swis said:

It won't matter. The problem is that code-wise, 'wheels' aren't actually round in the game, despite what your eyes tell you.

Ignore the round spinning visible part of the wheel that fools you into certain expectations, that is purely cosmetic. The actual wheel is an invisible skeleton without spinning parts so to speak that detects a nearby surface and creates a magical force under certain conditions. The code requires that the angle is close to perpendicular to the surface, or the 'wheel' is considered to not be touching and thus no force can be generated... despite the visual wheel spinning and touching or even clipping the surface and for all intents and purposes looking like it should be carving a ditch if it doesn't start moving soon.

 

Now cue the actual coders coming in to bap me upside the head and reassure everyone that Santa does actually exist, and wheels are in fact, really round and do spin, and we should all go back to sleep and all will be alright with v1.2 in the morning...
 

Yes I know this, the wheels are ray cast, not a round object. My point still stands, if they are not rotated past 45 degrees they will still work but wont move as each set of wheels are rotating in opposite directions. Reversing the drive of one set will allow for normal use of the wheels. Alas anything more than that and they wont apply drive force.

[NEBx]

1 hour ago, Majorjim said:

If the wheels are less than around 45 degrees they are spinning in different directions. Let me knowing reversing the drive of one pair of wheels helps. It should. 

If you notice in my original post that I tested anywhere from one to all axles powered, so this is not the issue.

What your saying is correct in their behaviour but I accounted for this in my testing.

Edited April 20, 2016 by NEBx

[sal_vager]

Trouble is, it only looks like a wheel, it's not actually a wheel, here, this pic should help.

[NEBx]

13 minutes ago, swjr-swis said:

It won't matter. The problem is that code-wise, 'wheels' aren't actually round in the game, despite what your eyes tell you.

Ignore the round spinning visible part of the wheel that fools you into certain expectations, that is purely cosmetic. The actual wheel is an invisible skeleton without spinning parts so to speak that detects a nearby surface and creates a magical force under certain conditions. The code requires that the angle is close to perpendicular to the surface, or the 'wheel' is considered to not be touching and thus no force can be generated... despite the visual wheel spinning and touching or even clipping the surface and for all intents and purposes looking like it should be carving a ditch if it doesn't start moving soon.

 

Now cue the actual coders coming in to bap me upside the head and reassure everyone that Santa does actually exist, and wheels are in fact, really round and do spin, and we should all go back to sleep and all will be alright with v1.2 in the morning...
 

I feared this was the case, but can it be rectified?

I'm no coder...

[Majorjim!]

21 minutes ago, sal_vager said:

Trouble is, it only looks like a wheel, it's not actually a wheel, here, this pic should help.

 

Yup, as was said above, its a raycast not a wheel.

23 minutes ago, NEBx said:

Ah ok I see, you tested them from 45 and over, my bad.

21 minutes ago, sal_vager said:

Trouble is, it only looks like a wheel, it's not actually a wheel, here, this pic should help.

 

Could the ray cast be set to always look 'down'?

[sal_vager]

11 minutes ago, Majorjim said:

 

Could the ray cast be set to always look 'down'?

No idea actually.

[swjr-swis]

17 minutes ago, Majorjim said:

My point still stands, if they are not rotated past 45 degrees they will still work but wont move as each set of wheels are rotating in opposite directions. Reversing the drive of one set will allow for normal use of the wheels.

(1) Rotated at exactly 45 degrees, they already sink into the tarmac and won't exert any force despite spinning. They will only 'grip' when trying to turn, and only the outside wheels, because the angle becomes just under 45 degrees - the 'inside' wheels get a higher angle so they still get no grip.

(2) As for the reverse spinning: this is due to the algorithm that makes wheels decide, based on the orientation of the command pod or probe core and their own angle compared to that, which way is 'forward'. The wheels that decide to inverse their spinning also depends on how the rover is built. In two quick rovers I built with a properly placed probe core facing up in the forward direction (the default SPH orientation), the left/right of each pair decided to spin in opposite direction, and due to the symmetry, they will switch but still spin opposite of each other when you 'reverse' them.

But due to (1), even when both left spun one direction and both right to the other, the rover would not rotate on its axis as one would expect, it would just sit there immobile, because no grip (unless turning the wheels).

 

(3) Now here comes the fun part in the 45 degree corner case: when you turn to one side (hold A or D), due to the way the wheel angle changes slightly relative to the pod/core and the way that affects spinning direction, one of the wheels on one side will now reverse (!), regardless of how you had it set to 'correct' this, meaning that at full turn the 'grippy' wheels will now work against each other. To see this clearly, hold A/D, brake, release brake, then tap W/S shortly to give it a slow spin. You'll notice one of the wheels is no longer spinning in the same direction. It sometimes doesn't happen at the first try, but just A/D W/S and brake a few times, move just the tiniest bit, and it will. Works easier with the M1, but also happens with the TR-2L.

Tested and confirmed again just now, with the TR-2L and M1 wheels, pure stock. 45 degrees is already too much; even a few degrees short of 45 will cause mayhem due to reversal when wheels are fully turned. Looks like one has to stay within ~40 degrees off perpendicular to be somewhat sure of grip and spin working as expected the whole time.

[swjr-swis]

1 hour ago, Majorjim said:

Could the ray cast be set to always look 'down'?

I think the problem with that is, what do we define as 'down'? It seems logical as long as the rover is on the flat runway with wheels mounted in the default orientation, but as soon as you start thinking of slopes and bumps... or those enterprising engineers who like to have wheels produce a force against something other than the ground, like other parts, of various (collider) shapes and at various angles, and in directions that do not run parallel to the planet's sea level or the ground.

So simply perpendicular to the terrain won't always work. Nor will the direction of gravity.

[Majorjim!]

6 hours ago, swjr-swis said:

I think the problem with that is, what do we define as 'down'? It seems logical as long as the rover is on the flat runway with wheels mounted in the default orientation, but as soon as you start thinking of slopes and bumps... or those enterprising engineers who like to have wheels produce a force against something other than the ground, like other parts, of various (collider) shapes and at various angles, and in directions that do not run parallel to the planet's sea level or the ground.

So simply perpendicular to the terrain won't always work. Nor will the direction of gravity.

Thank you for exposing my lack of thought on that issue!

 I guess it would be too expensive to have more than one ray caster?

[swjr-swis]

2 hours ago, Majorjim said:

I guess it would be too expensive to have more than one ray caster?

That's what I wonder too. I get that it's a computationally expensive thing, and we know KSP is already CPU bound as it is. But perhaps two casts instead of one would not be too much, say at an angle of 90 degrees from each other, 45 degrees away from the pure perpendicular. That would cover a 180 degree area. Half a wheel so to speak. Actually more like 165 or so, to avoid the edge case shenanigans, but close enough.

Still a good bit more than now, probably covering almost all use cases. At twice the CPU expense (per wheel). Worth a try maybe. Ideally as a setting, maybe defaulting to 1 as it is now, but allowing to set to 2 for the weirdos under us that like to angle our wheels now and then, or just to test the performance impact without affecting everyone else.

It may not be possible at all, if it's something inherited from Unity and the way it works is just how it is.

[Majorjim!]

1 hour ago, swjr-swis said:

That's what I wonder too. I get that it's a computationally expensive thing, and we know KSP is already CPU bound as it is. But perhaps two casts instead of one would not be too much, say at an angle of 90 degrees from each other, 45 degrees away from the pure perpendicular. That would cover a 180 degree area. Half a wheel so to speak. Actually more like 165 or so, to avoid the edge case shenanigans, but close enough.

Still a good bit more than now, probably covering almost all use cases. At twice the CPU expense (per wheel). Worth a try maybe. Ideally as a setting, maybe defaulting to 1 as it is now, but allowing to set to 2 for the weirdos under us that like to angle our wheels now and then, or just to test the performance impact without affecting everyone else.

It may not be possible at all, if it's something inherited from Unity and the way it works is just how it is.

Sorry to summon you again @Arsonide but do you know if this would be possible or at all practical?

[katateochi]

20 hours ago, Majorjim said:

Could the ray cast be set to always look 'down'?

As @swjr-swis points out, "what is down" is a problem. But could it not be something that we could adjust in the editor; ie have a tweakable that goes from 0 to 360 (0 being the "down" relative to the wheel mounting that it currently is set to).  That would give more flexibility for angled wheels. Still won't be able to build my Rollcage style car that can drive either way up, but I've come to accept that. 
Also yeah, what about more ray casters? I wonder how much extra load that would add. 

[NEBx]

Brilliant idea @katateochi!

That would not only solve design issues but would effectively remove any need for more rods too wouldn't it?

Hopefully it can be done

[katateochi]

22 hours ago, Majorjim said:

Sorry to summon you again @Arsonide but do you know if this would be possible or at all practical?

I also cast a spell of "sorry to bug you but, I summon @Arsonide" - can you also tell us if my suggestion of having a user adjustable angle for the wheel raycast is possible? (couple posts above) Thanks!

 

edit;

13 hours ago, NEBx said:

That would not only solve design issues but would effectively remove any need for more rods too wouldn't it?

And if it could be changed during flight via the menu or even better via action group (ie flip 180 degrees) then this kind of Rollcage inspired vehicle would be possible!

(sure, you'd have to drive backwards to go forwards when upside down, but I could live with that!)

But more seriously, if you could adjust the angle of raycast while in flight, you could adjust it to be more forward facing on the front wheels and have rovers more capable of climbing over obstacles.

Edited April 22, 2016 by katateochi

[swjr-swis]

1 hour ago, katateochi said:

But more seriously, if you could adjust the angle of raycast while in flight, you could adjust it to be more forward facing on the front wheels and have rovers more capable of climbing over obstacles.

If we're at the point of wishing for the impossible : worth considering whether the angle could change dynamically, and angle into the direction of movement, and back when at decelerating. That way when wheels are moving, they would automatically 'prepare' to deal with upcoming bumps/slopes, depending on speed.

[Majorjim!]

I just a horrible thought.. Legs are now wheels right? So they are subject to the same issues. I had legs set at an angle of more than 45 degrees in my horizontal Constellation landers, which worked perfectly in 1.05. They now either break or vibrate so much they kill the craft.. do legs 'deactivate' when set at more than 45 degrees too?? @Arsonide do you know if this is the case?

[swjr-swis]

3 hours ago, Majorjim said:

I just a horrible thought.. Legs are now wheels right? So they are subject to the same issues. I had legs set at an angle of more than 45 degrees in my horizontal Constellation landers, which worked perfectly in 1.05. They now either break or vibrate so much they kill the craft.. do legs 'deactivate' when set at more than 45 degrees too??

I decided to test this. Apparently, legs work at all angles; or at least, I couldn't get them angled enough to sink into the ground.

I built something really simple: a HECS core and the smallest LT-05 legs, placed in symmetry but rotated at 45 degree angle. It seems to rest on the legs like that without problem. Since the core's reaction wheels are strong enough to lift the whole thing on a single leg and keep it balanced like that, I tipped it over pretty much all the way where the top of the core was just inches over the ground... and it was still not falling over or sinking.

So, legs and wheels are obviously not handled entirely the same.

Obligatory screenshots:

 

[swjr-swis]

Hmm, interesting. I tested with the LT-10 next (LT-2 were too heavy for this), and observed two things:

• Not all legs are created equal. For whatever reason, the feet of the LT-10 sink quite a distance into the ground, where the LT-05 do not. They still show the same collision behaviour though, in that they seem unaffected by the angle at which they touch the ground.
• While trying different angles, I notice that the feet of the legs reorient themselves: when they are not 'touching ground', they go back to a default angle, but when they are touching (which for some reason is a good bit under the ground), they reorient themselves parallel to the terrain.

I wonder if when they had to rescale them, they forgot to check the collider scale, and it's still at the previous smaller size? But more importantly: it would seem that there is already code in place that checks the angle of the surface and adapts to it. I wonder if it's Squad's code, and whether it's reusable for the wheels...

 

 

Edited April 22, 2016 by swjr-swis
durn typos, and pic