Suggestion: Make wheels act like wheels instead of round sleds.

[Talavar]

  Redownloaded KSP after months of waiting for real wheel fixes.. made aircraft... skated around KSC for 2 minutes. Played with wheel traction.. skated more..  Uninstalled... it's Still crap. We don't need wheels that slide sideways. You may now all get mad at me for being irritated. Enjoy. (edit) Was playing 1.2

Edited September 16, 2016 by Talavar
Added info

[Stargate525]

Pick a subform. Don't cross-post.

Wheels skid. Trust me, it's MUCH better than when we first got them, as any deflection to them whatsoever made em dig in instantaneously. Next time try saying something constructive.

[TwinKerbal]

They are fixing wheels in 1.2

[steve_v]

2 hours ago, Talavar said:

Suggestion: Make wheels act like wheels instead of round sleds

They may be sleds, but they sure aren't round.

50 minutes ago, TwinKerbal said:

They are fixing wheels in 1.2

I'll call wheels 'fixed' when they actually have round colliders rather than points, and work with >15° caster. As far as I am aware the "wheels" in 1.2 are still ⤓, not ⍜.
Wheels should work when mounted sideways or upside-down too - wheels are supposed to be circular, and circles don't have an up and down.

This goes nowhere, because those are not wheels. Wheels are round.

Wheel: A circular object that revolves on an axle and is fixed below a vehicle or other object to enable it to move over the ground.

---

It will quite happily slide around sideways under reaction wheel torque, however. Perhaps not so far off with the 'sled' bit.

Edited September 16, 2016 by steve_v

[Speeding Mullet]

To bring my point across from the thread now closed due to duplication:

Which serves for the vast majority or use cases that KSP demands.  Don't get me wrong - I'm with you that they are not perfect, but they are a very, very long way from completely broken - I circumnavigated Minmus in a twenty wheeled land train this year with zero issues for example.  I also think you are oversimplifying the argument a little to suggest that it's as easy as fix sim, or build a better unity plugin.  I would suspect but cannot confirm that with the heat SQUAD has had on wheels that if this was a realistic prospect they would have done it already...

Regards your example picture above - Just mount the "round sleds" the right way up, and you'll be on your way with minimal aesthetic impact .

SM

 

[steve_v]

1 hour ago, Speeding Mullet said:

Just mount the "round sleds" the right way up, and you'll be on your way with minimal aesthetic impact

True, that is an exaggerated example.
That said, I have had aircraft catapaulted into the air on hitting a slight change in gradient many times - cockpit parts have a sloped underside resulting in nosegear castor, wheel hits uphill slope and stops dead then springs back.
Excessively energetic bouncy-bouncy when landing on rough terrain, same cause.
Perpetual-motion bounce-o-matic aircraft? The runway jitterbug? You guessed it.

Never had a wheel stop dead at the bottom of the slope driving down off the runway? I have.
Tired of having to make sure your landing gear are perfectly perpendicular to the ground to avoid physics jank? I am.

1 hour ago, Speeding Mullet said:

if this was a realistic prospect they would have done it already

Probably true, though I have to wonder if all the time spent on band-aids for dodgy wheel colliders would have been better spent on fixing the fundamental issue.

Wheels in 1.2 seem better, but going by various comments I have seen (along the lines of "suspension settings must be precisely tuned to each other" & "wheels must be perpendicular to the terrain" ) and the explode-o-matic rover ("That's not nominal!") I saw on one of the pre-release streams, it's just better band-aids on the same old guff.
My own experiences with wheels in 1.2 are similar, and not much removed from 1.1.*, just somewhat less jumpy and much less fragile. Dare I say it, a better workaround - though I'll also say they're damn near indestructible now.

It would appear the story is much the same with landing legs - landing legs that kerbals can now walk right through.
AFAICT, this is the latest workaround for, wait for it... The single point collider reacting to an instant 'no kerbal under leg' -> 'kerbal under leg' transition, with explosive results.
Down with the over-simplified not-wheel not-landing leg single point wheel colliders, and the many problems they cause.

---

A very quick search turns up UWP, clearly it doesn't have everything KSP needs, but this kind of thing certainly can be done. Multiple raycasts or possibly sphere-casting would result in much better behaviour on rough terrain. A true 3D tyre would be even better, but one would have to watch the performance overhead.

Edited September 16, 2016 by steve_v

[Majorjim!]

Wheels are a single ray cast 90degrees down.. That is why a large wheel falls between a very narrow gap in panels and why you often tip over forward when transitioning from a slope to flat area. Yes this is less than perfect but KSP is not like a driving sim. You can, in KSP set wheels upside down and in all manner of crazy positions. Maybe all driving sims use ray cast wheels. I say it's good enough and issues are easy to make allowances for. 

[Arsonide]

20 minutes ago, steve_v said:

"suspension settings must be precisely tuned to each other" & "wheels must be perpendicular to the terrain" ) and the explode-o-matic rover ("That's not nominal!") I saw on one of the pre-release streams

Suspension settings in any vehicle, and any wheel solution would need to be precisely tuned to each other, whether or not the wheel was simulated with a raycast or a round volume of some sort. If you are referring to the rover that EJ blew up, that was an issue with the camera, not the wheels. He crashed upside down and blew up his root part, which triggered an issue we've resolved since with the camera.

5 hours ago, steve_v said:

'll call wheels 'fixed' when they actually have round colliders rather than points

The Unity wheel collider is a raycast pointing in the direction of gravity, rather than a point. For the majority of vehicles, players, and games that utilize this collider, this is fine. Most third party solutions for wheels are also raycasts pointing in the direction of gravity, with some extra features. If we were to change this paradigm, it would take a great deal of development time away from other things, to allow a very small subset of vessels to function.

Edited September 16, 2016 by Arsonide

[steve_v]

4 minutes ago, Arsonide said:

The Unity wheel collider is a raycast pointing in the direction of gravity, rather than a point.

So why not multiple raycasts radially from the centre of the wheel? I don't know Unity, but I don't see any reason for raycasting only in the direction of gravity.

[Majorjim!]

11 minutes ago, steve_v said:

So why not multiple raycasts radially from the centre of the wheel? I don't know Unity, but I don't see any reason for raycasting only in the direction of gravity.

I've suggested this before but it was ignored. 

[steve_v]

5 minutes ago, Majorjim! said:

I've suggested this before but it was ignored. 

I'd be surprised if someone hadn't, seems like an obvious solution to me. Of course one would have to roll your own ground-contact detection, suspension, friction, all the other stuff that comes with the existing wheel system... but I don't see why it couldn't be done.
I can only speculate that it comes down to "no off-the-shelf solution available, too much work to implement in-house".

Edited September 16, 2016 by steve_v

[katateochi]

22 minutes ago, Arsonide said:

The Unity wheel collider is a raycast pointing in the direction of gravity

If the raycast is pointing in the direction of gravity, wouldn't that allow for "upside down" or 90 degree rotated wheels to work?
Also, if pointing in the direction of gravity, how does that impact a rover going up a slope? Would that mean that on certain inclines the wheels become...not wheels, just because of the angle of slope?
I do think though that the wheels have been greatly improved in the 1.2 prerelease (I'm presuming that what this thread is in reference to).

 

18 minutes ago, steve_v said:

So why not multiple raycasts radially from the centre of the wheel

I'm guessing that each raycast adds a performance overhead, which might be why only one is used.  Just a guess though.

[Majorjim!]

4 minutes ago, katateochi said:

If the raycast is pointing in the direction of gravity, wouldn't that allow for "upside down" or 90 degree rotated wheels to work?
Also, if pointing in the direction of gravity, how does that impact a rover going up a slope? Would that mean that on certain inclines the wheels become...not wheels, just because of the angle of slope?
I do think though that the wheels have been greatly improved in the 1.2 prerelease (I'm presuming that what this thread is in reference to).

 

I'm guessing that each raycast adds a performance overhead, which might be why only one is used.  Just a guess though.

My wording was, I believe, more accurate. The raycast is 90degrees downward from the wheels upright position. I believe. ;-)

[Arsonide]

3 minutes ago, Majorjim! said:

My wording was, I believe, more accurate. The raycast is 90degrees downward from the wheels upright position. I believe. ;-)

Correct, we overload the default and use the wheel's down, given that our gravity can go in any direction.

[steve_v]

16 minutes ago, Arsonide said:

Correct, we overload the default and use the wheel's down, given that our gravity can go in any direction.

Workaround for driving on steep terrain and flying space-wheels, breaks upside-down operation. Still rather sub-optimal IMO.
Again, why not several radial raycasts? Would this not potentially work for all situations?

Edit: I'll just drop this here, 3D tyres, apparently. Certainly seem to behave better that the ones we have:
https://www.youtube.com/watch?v=hbaPUZDJpl8

 

Edited September 16, 2016 by steve_v

[The_Rocketeer]

25 minutes ago, Arsonide said:

Correct, we overload the default and use the wheel's down, given that our gravity can go in any direction.

Why does this seem to affect 'pitched' wheels more than 'cambered' wheels?

[Arsonide]

23 minutes ago, steve_v said:

Edit: I'll just drop this here, 3D tyres, apparently. Certainly seem to behave better that the ones we have:
https://www.youtube.com/watch?v=hbaPUZDJpl8

This video was uploaded in 2011. The package itself was last updated in 2012, and therefore has not had to deal with all of the...intricacies...of Unity 5's wheel colliders. It would no longer function nowadays.

[katateochi]

2 minutes ago, Majorjim! said:

My wording was, I believe, more accurate. The raycast is 90degrees downward from the wheels upright position. I believe. ;-)

That does seem to fit the observed behaviour better as a 90 degree (or IIRC > 25 degree) rotated wheel no longer functions.

 

1 minute ago, Arsonide said:

Correct, we overload the default and use the wheel's down, given that our gravity can go in any direction.

ah...interesting. But within an SOI isn't gravity always towards the center of the current planet? 
Having the raycast pointing towards the center of the current SOI would seem to solve some of the problems, ie allow wheels to be placed at any rotation and probably stop the case of a wheel getting stuck when coming down a steep slope.  But it also seems like it would cause problems, ie loss of functionality when going up a steep slope, no longer usable as bearings and of course that would prevent folk building trackmania style loop the loop courses for rovers. 
So neither way is ideal. 

But what about multiple raycasts as @steve_v suggests? Cos that really sounds like the best solution. Does that just add too much overhead?

[steve_v]

Just now, Arsonide said:

It would no longer function nowadays.

True, I wasn't suggesting you switch to that particular package. My point was to show that round wheels that react to terrain in a realistic manner are quite possible, and to illustrate just how hard single-raycast wheels suck in comparison.
 

[Sharpy]

I have to say one thing: Considering the DISASTER the wheels in 1.1.3 were, the 'performance impact' argument doesn't really hold water.

I mean, removing the wheels altogether would result even better performance, at no significant loss of functionality.

[klgraham1013]

Being no developer, I have also had unqualified thoughts on this.  Having a main raycast facing the center of the planet, 2 more at 45 degrees either side of the main raycast, and 2 at 90.  Sure it adds more over head, but so does making an 18 wheeler.  At least 4 wheelers might function better.

1 hour ago, Arsonide said:

If we were to change this paradigm, it would take a great deal of development time away from other things, to allow a very small subset of vessels to function.

Rovers are only a small subset because their use case is a lot smaller in KSP than it is in real life.  Also, as the game seems to be reaching some sort of final form (in my opinion, and not in a bad way), isn't this excuse of "taking away from other things" wearing thin?  I can't think of many more features (besides KER and life support) that actually could be; have the possibility of being added.  I love the polish you've put into 1.2, and would love to see this trend continue, perhaps encapsulating a hard look at the wheel code.

Thanks for the work you've done to make wheels usable in 1.2.

[steve_v]

Allow me to quickly demonstrate why the current wheel system is rubbish... want to build an all-terrain vehicle? Out of luck, you will hang up un the first brick (or any rectangular object) you encounter.

And no, it's not catching on the underside, wheels just can't handle rectangular obstacles. At all.

Real moon rovers, for example, are specifically designed with wheels that can ride over small rocks and climb acute angles... maybe this is why terrain scatter is still not solid - those small stones would likely stop a KSP wheel dead.

Edited September 16, 2016 by steve_v

[Arsonide]

26 minutes ago, klgraham1013 said:

isn't this excuse of "taking away from other things" wearing thin?........I love the polish you've put into 1.2

Unfortunately these two statements directly contradict each other. Polish does not come from the polish fairy. Every second we spend on Feature X is a second that Feature Y does not get.

[Talavar]

8 hours ago, TwinKerbal said:

They are fixing wheels in 1.2

 I was playing 1.2 .. They are still broke.  Something simple, like taking off in an aircraft, turns into Disney on ice. You cant stay on the runway long enough, without kicking sideways, to take off in an aircraft. Drag was behind center of mass, center of thrust balanced with center of mass, No reason it shouldn't fly. 1.0.5 may not have been perfect, but at-least wheels functioned. I'm just thinking, perhaps, adding slip to the tires wasn't such a good idea... Yeah, they tipped easier before, but at least you could take off in an air-craft

[steve_v]

16 minutes ago, Talavar said:

Something simple, like taking off in an aircraft, turns into Disney on ice. You cant stay on the runway long enough, without kicking sideways, to take off in an aircraft.

Well, I just flew 2 different SSTOs (mixture of small & medium retractable gear) to orbit and back, takeoff from the runway, landed on some random lumpy grass, no real problems. The gear still slide sideways more than they should IMO, but not exactly "Disney on ice" and quite manageable really. The demon possessed pogo stick effect on touchdown is also much reduced.

Which gear are you using anyway? Takeoff speed? Craft Mass & gear layout?
Pics? demo vid? proof? You know the old saying...

Edited September 16, 2016 by steve_v