Euler/Quaternion conversion issue

[lostwoods91]

hi everyone!

what's wrong with this?: http://playground.babylonjs.com/#C0VAG#2

first i set an orientation vector, called eulAngles, then i apply it to a cube.

after conversion to a quaternion and viceversa, the angles are slightly wrong!

maybe i'm doing something wrong (not sure if it is a bug), but it's strange...

the conversion is correct with these rotations: X, Y, Z, XY, YZ; it is wrong with: XZ, XYZ (general case) rotations instead.

so i noticed that it is wrong in all cases X and Z are rotated together...

i've already tried switching axis orders, but nothing!

[hen]

I was dealing with the same problem, here is a a quaternoin conversion that works:

quaternionToEuler(q) {
        var heading, attitude, bank;
        var x = q[0], y = q[1], z = q[2], w = q[3];

        var test = x*y + z*w;
        if (test > 0.499) { 
            heading = 2 * Math.atan2(x,w);
            attitude = Math.PI/2;
            bank = 0;
        }
        if (test < -0.499) { 
            heading = -2 * Math.atan2(x,w);
            attitude = - Math.PI/2;
            bank = 0;
        }
        if(isNaN(heading)){
            var sqx = x*x;
            var sqy = y*y;
            var sqz = z*z;
            heading = Math.atan2(2*y*w - 2*x*z , 1 - 2*sqy - 2*sqz);
            attitude = Math.asin(2*test);
            bank = Math.atan2(2*x*w - 2*y*z , 1 - 2*sqx - 2*sqz);
        }

        return [bank, heading, attitude];
    };

 

[lostwoods91]

thank you! i'll try it soon

[adam]

It appears that BJS does rotation in the order ZXY.  That toEulers function only supports YZX.

https://github.com/BabylonJS/Babylon.js/blob/master/src/Math/babylon.math.ts#L1916

https://github.com/BabylonJS/Babylon.js/blob/master/src/Math/babylon.math.ts#L1938

[hen]

19 minutes ago, adam said:

It appears that BJS does rotation in the order ZXY.  That toEulers function only supports YZX.

https://github.com/BabylonJS/Babylon.js/blob/master/src/Math/babylon.math.ts#L1916

https://github.com/BabylonJS/Babylon.js/blob/master/src/Math/babylon.math.ts#L1938

I am using this function alot in our app.. it works, guaranteed

[adam]

Can someone explain to me why there seems to be a difference in rotation order between what BJS does and what the Quaternion.toEuler function appears to support?

[lostwoods91]

21 minutes ago, hen said:

I am using this function alot in our app.. it works, guaranteed

How do you explain my playground demo, then? How can I fix it? 

Thanks 

[hen]

On 10/12/2016 at 9:36 AM, lostwoods91 said:

How do you explain my playground demo, then? How can I fix it? 

Thanks 

Just try this: http://playground.babylonjs.com/#C0VAG#3

[adam]

23 minutes ago, adam said:

Can someone explain to me why there seems to be a difference in rotation order between what BJS does and what the Quaternion.toEuler function appears to support?

I'm pretty sure I was thinking in terms of world space and I should have been thinking in local space.

[adam]

On 10/12/2016 at 9:45 AM, hen said:

Just try this: http://playground.babylonjs.com/#C0VAG#3

Same problem:

http://playground.babylonjs.com/#C0VAG#4

[hen]

Weird... i must say that im still using v2.4 here.
No idea what could be wrong, maybe the problem is somewhere else.

[lostwoods91]

Maybe you use it without having X and Z axis together in the same rotation...? 

[JohnK]

Rotations are very tricky, not least because they are non-commutative and can take place in WORLD or LOCAL space. Also the toEulerAngles function returns the angles in the order  YZX not XYZ

In BJS the use of mesh.rotation is in WORLD space and yaw, pitch, roll take place in local space.

In the example http://playground.babylonjs.com/#C0VAG#6

mesh.rotate is used and this can be set to use LOCAL space. The box is taken to be moving in the positive z direction with  roll about z axis, pitch about x axis  and yaw about y axis. The rotations are first done in the order yaw, pitch, roll. To find the quaternion the prameters must be entered in yaw, pitch, roll order. The function toEulerAngles returns yaw, roll, pitch and the final rotations are applied in this order.

In this example http://playground.babylonjs.com/#C0VAG#7

The box is taken to be moving in the positive y direction so roll about y axis, pitch about x and yaw about z

The rotation is applied to cube1 and cube 2 shows that this matches with rotate in the order ZXY in WORLD space.

The values for eulAngles in the order yaw, pitch roll, ie zxy, are used to determine the quaternion and apply it to cube3.

The euler angles are then found using toEulerAngles in the order YZX and are applied locally in that order to give the same rotation to cube4. 

[adam]

I think I figured it out:

http://playground.babylonjs.com/#C0VAG#10

I used the function from this page:

https://en.wikipedia.org/wiki/Conversion_between_quaternions_and_Euler_angles

I had to modify it so that the rotation order is ZXY.

[GameMonetize]

Sorray late to the party:) is there a bug in our math lib?

[adam]

There does appear to be a bug in Quaternion.toEulerAnglesToRef.

[JohnK]

Looking at Quaternion.toEulerAnglesToRef in Babylon max it takes a parameter - a string that was intended to provide the return order. It seems it was never completed so just uses the default order.

[GameMonetize]

Happy to validate a PR here  (not enough hours in a single day to do everything  )

[adam]

I'll submit a PR this weekend.

[DigiHz Data]

1 hour ago, adam said:

I'll submit a PR this weekend.

Plz do that @adam .I am trying to implement @satguru's edit control in my editor, and i have problems with rotation values getting back right (using euler values).

I think the function Quaternion.prototype.toEulerAnglesToRef = function (result, order) {......needs to be updated with  support for  ZXY  as well.

Maybe other orders to???

As @JohnK mentioned......the function is not complete yet. I would not call it a bug, but missing options about the order.

[adam]

5 minutes ago, DigiHz Data said:

I think the function Quaternion.prototype.toEulerAnglesToRef = function (result, order) {......needs to be updated with  support for  ZXY  as well.

Can someone prove that the current conversion is working correctly (YZX order)?

 

[DigiHz Data]

On 10/13/2016 at 9:17 AM, adam said:

I think I figured it out:

http://playground.babylonjs.com/#C0VAG#10

I used the function from this page:

https://en.wikipedia.org/wiki/Conversion_between_quaternions_and_Euler_angles

I had to modify it so that the rotation order is ZXY.

Can you give me the modificated code for ZXY ?

[adam]

The quatToRot function seems to be working for both rotation orders.  I tested it in world and local space too.

http://www.babylonjs-playground.com/#1FZQZH#7

Here is another test with random rotation:

http://www.babylonjs-playground.com/#1FZQZH#8

I don't think it's going to be necessary to specify the order of rotation when converting from quat to euler.

 

[adam]

I just submitted the pull request.

This PG should work correctly once the PR is validated and the PG is updated.

http://www.babylonjs-playground.com/#1FZQZH#10

[GameMonetize]

PG updated

PG works

Thanks a lot @adam!