Rovers not running quite straight.

[MarvinKitFox]

Why do all of my rovers turn to the side, all the time?

Not a violent swerve, mind you, but a constant very gradual creep off to one side.

The direction of turn tends to be consistent for a given model in a given situation, but this is not 100%

Even on perfectly level ground (minmus's flats are as....flat a terrain as one could ask for.)

The rover is designed using 100% 2-way mirrored parts, so no possible question of it being asymmetrical in dimensions, shape, mass, or thrust.

I've tried building in VAB and Hanger, no change.

I've made sure to eliminate ground slope. (flats on minmus)

I've made sure I have ample power, so no one wheel will get starved, ofsetting the thrust.

I've even considered coriolis force as the imbalance. I made sure my MinMus testing run was *exactly* east-west along the equator. (+- less than 1m deviation in a 5km run)

Ive tried all the various wheel types, no luck.

What am I doing wrong?

Edited March 6, 2014 by MarvinKitFox

[Wooks]

I´ve consistently found this behaviour even in my most balanced rovers, for example take a look at this timelapse of my first passing over the North Pole:

I´m driving straight pressing only W to move the rover forward, and I have to make constant course corrections because it wants to turn East constantly. Why? Beats me.

[Kasuha]

Even on Minmus flats you're still traveling on a ball. A completely straight trajectory on a ball is a circle with headings e.g. 45° on equator, 90° once you reach your norhtest latitude, 135° when you're back on equator, and 90° again when you reach your southest latitude. The only four directions which don't change is directly North, South (they still do, but abruptly on the pole), East, or West (these two truly don't change).

Edited March 5, 2014 by Kasuha

[Wooks]

The only four directions which don't change is directly North, South (they still do, but abruptly on the pole), East, or West (these two truly don't change).ht

True for the abrupt change over the pole.

[MarvinKitFox]

Even on Minmus flats you're still traveling on a ball. A completely straight trajectory on a ball is a circle with headings e.g. 45° on equator, 90° once you reach your norhtest latitude, 135° when you're back on equator, and 90° again when you reach your southest latitude. The only four directions which don't change is directly North, South (they still do, but abruptly on the pole), East, or West (these two truly don't change).

Yep.

That's why I ran my ultimate test on Minmus Flats.

+Perfectly level.

+Exactly east-west

+Exactly on equator.

My rover still turned right by about 1 degree for every 100m of distance. :9

[WhiteKnuckle]

Have you tried turning off the torque from your probe core/command pod + any reaction wheels?

[Kasuha]

I remember driving my rover for about 30 km in straight line on Laythe's poles and I don't remember any such issues. It was rather far so I was dealing with it using an "autopilot" paper wedged to my keyboard so I guess I would sure notice it is turning. That was in 0.22 though.

Things that come on mind:

- if your rover is low on electricity, one of wheels may be generating less power (similar to uneven thrust in multi-jet plane designs)

- if your rover is unbalanced, more weight on one side may be causing this

[Vanamonde]

Two things: 1) If one side of your vehicle is getting a little more traction, it will try to pull ahead of the other side, resulting in the vehicle turning toward the slower side. The weird thing is, this can happen even when the ground *appears* to be perfectly flat. 2) Is the vehicle heavy enough that the structure sags? If so, your wheel mounts might be twisting and therefore no longer perpendicular to the ground. That will also cause swerving, to varying degrees.

[little square dot]

You're right; I don't believe that you loved every minute of it.

=)

[celem]

I suspect Vanamonde is probably in the right area here. A lot of my designs visibly sag a little which of-course leads to non-straight non-perpendicular wheels.

I've certainly noticed this too and have never really succeeded in eliminating it from designs.

In order to spread the weight as evenly as possible I started using a great many wheels on each side in order to remove suspension compression as a possibility, still turns.

Tried the east-west minmus flats, still turning.

Tried killing all the torque modules and driving in docking mode, still turns.

What really weirds me out, is that if Vanamonde identified the only factor here, then the direction of the drift wouldnt be consistent, sometimes you would drift left, sometimes right, depending on the design and surface. I always see left-drift...

Edited March 6, 2014 by celem

[Wooks]

Couple things I noticed in my experience over the pole:

1.- At normal time warp my rover does not turn east. AT ALL. It just keeps trucking merrily along the path.

2.- Most of my trip is made under 3x and 4x time warp because I don´t want it to take forever, but then the rover wants to turn East every time.

Again, not a big deal, rover is perfectly balanced (it seems to be) and it doesn´t seem to affect performance.

You're right; I don't believe that you loved every minute of it.

=)

I am SO easily amused

[MarvinKitFox]

Two things: 1) If one side of your vehicle is getting a little more traction, it will try to pull ahead of the other side, resulting in the vehicle turning toward the slower side. The weird thing is, this can happen even when the ground *appears* to be perfectly flat. 2) Is the vehicle heavy enough that the structure sags? If so, your wheel mounts might be twisting and therefore no longer perpendicular to the ground. That will also cause swerving, to varying degrees.

Umm.... I would have though MinMus's grabbity is feeble enough, but yes... My rovers are working machines, they carry lots of stuff. The one I was testing on is about 2.7T... Thats (2700*0.491/9.81/4)=33.8kg per wheel... thats not going to bend anything, much less the Modular Girder Segment XL it is directly attached to.

Now on my Shongololo diesel-electric Explorer, the wheel loading is a bit more. about 11.3tons per wheel And yes, it crawled, crabbed and zigzagged most enthusiastically, due to very visible frame deformations.

Meh, the problem is always addressable with either a spot of trim, or using mechjeb's (pathetic!) rover automation.

[little square dot]

Considering that most people have this issue, and considering that I've had better luck with mod wheels and the RBI tracks, is it possible that the CoM is slightly off in the part itself?

ie. if the wheel part's CoM is slightly forwards or backwards of centre your vehicle would always deviate slightly, as the wheels are the inverse of one another meaning that one wheel's CoM would be offset forwards and the other one backwards, thus the pulling to the left or right.

[Geschosskopf]

Considering that most people have this issue, and considering that I've had better luck with mod wheels and the RBI tracks, is it possible that the CoM is slightly off in the part itself?

It seems to me more like KSP doesn't attach parts quite straight. If you zoom in really close, you can see this. Also happens with when using a winglet as a vertical tail / rudder on an airplane, so the airplane also slowly drifts to the right.

[Xiphos]

This is why all my rovers still use the 'landing gear' wheels. I have the front ones set to raise, allowing standard rover wheels to make contact when sharp turns need to be made. Also no wheels currently can haul as much a$$ as these can. heh.

[little square dot]

I am SO easily amused

hahaha... evidently.

It seems to me more like KSP doesn't attach parts quite straight.

That was my first thought too, but after checking the .craft files of several vehicles, it would seem that the positions of symmetrically placed wheels are identical, save the inverted x-axis.

I probably should have assumed this to be the case given that it would actually be more difficult to code the game to be imprecise than to simply invert the x-axis and rotate the y-axis 180 degrees.

I think the two most likely explanations would have to be either flex in the vehicle or a slight imprecision in the part itself. I'll just weld a vehicle into a single part to test this. Given that the .craft shows identical coordinates on both sides, if the welded vehicle deviates from the centre-line of the runway, it would have to be a flaw in the part... or the runway.

Edited March 6, 2014 by little square dot

[little square dot]

Just tested the wheelMed. Rover body is a single welded part and according to the .craft file the wheels are perfectly symmetrical.

The results:

[little square dot]

ok, I am officially at a loss.

This is just bizarre... didn't change a thing; simply reverted to launch.

Test #2 -much less deviation

Test #3 -deviates in the opposite direction!

[little square dot]

btw, Wooks, I had a hard time just sitting through the drive to the end of the runway 3 times, haha..

[Xaiier]

Have none of you heard of floating point errors?

[vger]

I ran into this issue as well but I don't believe it is a problem with the design (some of the time anyhow). When I was tweaking a car for the F1 challenge, I found that sometimes it would drive straight, and other times it wouldn't. Something about the initial 'drop' at the launchpad seems to influence it, and it's more or less the luck of the draw whether the car winds up perfect. Sometimes I noticed even the spoiler would be crooked (blatantly so) and other times it would be straight. The only way to fix it was by reverting to launch a few times until it worked.

[little square dot]

Well if that's the explanation, clearly no.

Would you care to enlighten us, or are you merely here to highlight our ignorance?

nm, I'll fire up google.

[Kasuha]

Maybe you could post your craft file so we can check it and try if there's something wrong.

[little square dot]

Ah, computers suck at maths.

http://www.techradar.com/news/computing/why-computers-suck-at-maths-644771/1

(thanks for the heads-up=)

[Xaiier]

Ah, computers suck at maths.

http://www.techradar.com/news/computing/why-computers-suck-at-maths-644771/1

(thanks for the heads-up=)

Unity uses PhysX, which uses floats for all its calculations. It can be easily seen when doing some basic math tests that float variables are terrible at accuracy when dealing with big numbers, and slight random variation from the true values adds up to a small curve over the course of a drive. This is the same reasoning behind why the kraken ate ships before krakensbane was implemented, though in this case its on the smaller end of the scale rather than the large one.