Jump to content

  • Log In with Google      Sign In   
  • Create Account


Rotate 4x4 matrix formula


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
4 replies to this topic

#1 Krik   Members   -  Reputation: 125

Like
0Likes
Like

Posted 22 February 2013 - 07:41 PM

I am trying to figure out how to work with a 4x4 matrix, in particular rotating them. I have been searching and reading differnt things but I keep hitting a brick wall when I see the formulas. I can see which parts of the matrix need to be changed depending on if I rotate on the x, y or z axis. But I cannot interpet their formulas so as to use them in code.

 

I have tried several experiments and I am not geting the right numbers. And since they use letters and greek letters and give no point of refernce for them I am sure additional calculation are need but there is no clear reference to the correct calculation. In one fomula I saw they used cosine and sine with the theta character in parenthases which previously in the article it was defined as pi/2 (which makes no sense at all) but that didn't work when I used it.

 

I am hoping someone has some nice clearly defined formulas I can work with.

 



Sponsor:

#2 Geometrian   Crossbones+   -  Reputation: 1503

Like
0Likes
Like

Posted 22 February 2013 - 11:37 PM

A piece of code is worth a thousand words. The following comes from my C++ math library. It's templated, but it should be fairly readable. It takes a unit vector representing the axis to rotate around and an angle in radians and returns a rotation matrix. See also the entries on Wikipedia.

//Creates a rotation matrix from radians
template <typename type_elements, typename type_float1, typename type_float2>
Matrix<type_elements,4,4> get_mat4_rotation_rad(const type_float1 vec[3], type_float2 theta) {
    type_elements c = cos( (type_elements)(theta) );
    type_elements s = sin( (type_elements)(theta) );
    type_elements C = (type_elements)(1) - c;
    type_elements x   = vec[0], y   = vec[1], z   = vec[2];
    type_elements xs  =   x* s, ys  =   y* s, zs  =   z* s;
    type_elements xC  =   x* C, yC  =   y* C, zC  =   z* C;
    type_elements xyC =   x*yC, yzC =   y*zC, zxC =   z*xC;
    return Matrix<type_elements,4,4>(
        (double)(x*xC+c),(double)(xyC-zs),(double)(zxC+ys),0.0,
        (double)(xyC+zs),(double)(y*yC+c),(double)(yzC-xs),0.0,
        (double)(zxC-ys),(double)(yzC+xs),(double)(z*zC+c),0.0,
                     0.0,             0.0,             0.0,1.0
    );
}

-G


Edited by Geometrian, 22 February 2013 - 11:37 PM.

And a Unix user said rm -rf *.* and all was null and void...|There's no place like 127.0.0.1|The Application "Programmer" has unexpectedly quit. An error of type A.M. has occurred.

#3 Krik   Members   -  Reputation: 125

Like
0Likes
Like

Posted 23 February 2013 - 05:04 AM

Thanks Geometrian for posting that code. I haven't worked in C++ for about a year so I will have to take a bit of time to dig into that and see what I can glean from it.

Since my post I have found a few more examples and articles but I am still a bit off on my calculatios. I am sure some will laugh but the formula I found that cracked everything open was not a formula to create the matrix but a formula that allowed one to pull the degrees of rotation out of the matrix. I took it and just reversed it, then from there I knew where to place the results and I figured out matrix multiplicaion.

 

But like I said things are bit off. There are some built in functions for simplifing rotation (they are limiting what I can do), and I am comparing my numbers to them. If I just rotate on one axis the numbers are fine it is when I rotate on more than one that things get messed up.

 

I have my code outputing the numbers from the built in rotation fuctions (first matrix below) and from my calculations (second matrix below), here's what I am seeing

X: 2deg
Y: 2deg

0.9993908270190958, 0.001217974870087876, -0.03487823687206265, 0,
0, 0.9993908270190958, 0.03489949670250097, 0,
0.03489949670250097, -0.03487823687206265, 0.9987820251299122, 0,
99.93908270190957, 100.06088018891836, 0.002125983043832047, 1

X cosine: 0.9993908270190958
X sine: 0.03489949670250097
Y cosine: 0.9993908270190958
Y sine: 0.03489949670250097

0.9993908270190958, 0, -0.03489949670250097, 0,
0.001217974870087876, 0.9993908270190958, 0.03487823687206265, 0,
0.03487823687206265, -0.03489949670250097, 0.9987820251299122, 0,
100, 100, 0, 1

Note that there is a 100 translate on both the X and Y axis so I can throughtly test my numbers.

I looks like the X and Y cosine and sine calculations are correct but something is not being calculated correctly after that. Nor is the calculation being applies to the translate values (I suspect I need to just research a bit to solve this). Hmm, as I look a little closer it looks like maybe some numbers are in the wrong spot ... but why?

 

Here is a pseudo version of the code I currently have

basematrix = [
    [1, 0, 0, 0],
    [0, 1, 0, 0],
    [0, 0, 1, 0],
    [100, 100, 0, 1]
];

function RotateMatrix(y, x)
{
    halfrot = 3.141592653589793 / 180;	// pi divided by 180

    xcos = cos(x * halfrot);
    xsin = sin(x * halfrot);

    ycos = cos(y * halfrot);
    ysin = sin(y * halfrot);

    ymatrix = [
        [ycos, 0, -ysin, 0],
        [0, 1, 0, 0],
        [ysin, 0, ycos, 0], 
        [0, 0, 0, 1]
    ];

    xmatrix = [
        [1, 0, 0, 0],
        [0, xcos, xsin, 0],
        [0, -xsin, xcos, 0], 
        [0, 0, 0, 1]
    ];

    calcmatrix = MatrixMultiply(ymatrix, basematrix);

    calcmatrix = MatrixMultiply(xmatrix, calcmatrix);

    for (i = 0; i < 4; i++)
    {
        for (j = 0; j < 4; j++)
        {
            //output calcmatrix[i][j]
        }
    }
}

function MatrixMultiply(matrixa, matrixb)
{
    newmatrix = [];
    for (i = 0; i < 4; ++i)
    {
        for (j = 0; j < 4; ++j)
        {
            newmatrix[i][j] = matrixa[i][0] * matrixb[0][j] 
                            + matrixa[i][1] * matrixb[1][j] 
                            + matrixa[i][2] * matrixb[2][j] 
                            + matrixa[i][3] * matrixb[3][j];
        }
    }	
    return newmatrix;
}

Maybe someone can see what I did wrong.


Edited by Krik, 23 February 2013 - 05:08 AM.


#4 ankhd   Members   -  Reputation: 1242

Like
0Likes
Like

Posted 24 February 2013 - 07:26 AM

Hi.

Maybe some thing like this

float4x4 rotateX = { 1.0, 0.0, 0.0, 0.0,                0.0, cos(rot.x), -sin(rot.x), 0.0,                0.0, sin(rot.x), cos(rot.x), 0.0,                0.0, 0.0, 0.0, 1.0  };
float4x4 rotateY = { cos(rot.y), 0.0, sin(rot.y), 0.0,                0.0, 1.0, 0.0, 0.0,                -sin(rot.y), 0.0, cos(rot.y), 0.0,                0.0, 0.0, 0.0, 1.0   };
float4x4 rotateZ = { cos(rot.z), -sin(rot.z), 0.0, 0.0,                sin(rot.z), cos(rot.z), 0.0, 0.0,                0.0, 0.0, 1.0, 0.0,                0.0, 0.0, 0.0, 1.0   };



#5 stannic   Members   -  Reputation: 173

Like
0Likes
Like

Posted 24 February 2013 - 02:25 PM

My guess is that your problem has something to do with notation and order of operations.

 

 

Remember that  [A][X][Y] != [Y][X][A] != [X][Y][A] != [A][Y][X]






Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS