I have a sphere orbiting another sphere, (say a moon orbiting a planet) using glRotate. How do I determine the xyz positon of the moon at any particular point in its orbit? I need this info for frustum culling and lighting.

the glRotate man page tells you how the rotation matrix is constructed. Multiply your modeling transformation matrices by the object's position to find its final world position.

OneSadCookie Wrote:the glRotate man page tells you how the rotation matrix is constructed. Multiply your modeling transformation matrices by the object's position to find its final world position.

Thanks, OSC, you've pointed me in the right direction. Haven't quite gotten my arms around matrices just yet, but I'll keep digging. I was hoping you wouldn't use 'object's position' in your reply, because that's the unknown piece that I'm looking for.

Best regards,
charon

sorry, the object's original position.

basically, you need to reproduce yourself the steps that OpenGL will take in transforming your object.

check out Matricies can be your Friends

http://sjbaker.org/steve/omniv/matrices_...iends.html

And you can see that m[12], m[13], m[14] hold the x, y, z of the translation part. So you could do your rotations and translations normally - then read back the matrix from OpenGL (this is slow!) - and then read m12-14 to find out where the planet is.

// This is typed in Safari
// Its not real code
// so it could be wrong
// but I imagine doing something like
GLfloat transform_matrix[16];
glMatrixMode(GL_MODELVIEW);
glLoadIdentity();
//
// do some transformations - glRotate() and glTranslate() etc...
//
glGetFloatv (GL_MODELVIEW_MATRIX, transform_matrix); //slow!!
GLfloat planetXLocation = transform_matrix[12];


After you get this working - you can substitute your own transformation calls that mirror the OpenGL ones so you no longer have to pull the matrix back from the graphics card if its too slow.

hth,
Codemattic

codemattic Wrote:check out Matricies can be your Friends

http://sjbaker.org/steve/omniv/matrices_...iends.html

And you can see that m[12], m[13], m[14] hold the x, y, z of the translation part. So you could do your rotations and translations normally - then read back the matrix from OpenGL (this is slow!) - and then read m12-14 to find out where the planet is.

// This is typed in Safari
// Its not real code
// so it could be wrong
// but I imagine doing something like
GLfloat transform_matrix[16];
glMatrixMode(GL_MODELVIEW);
glLoadIdentity();
//
// do some transformations - glRotate() and glTranslate() etc...
//
glGetFloatv (GL_MODELVIEW_MATRIX, transform_matrix); //slow!!
GLfloat planetXLocation = transform_matrix[12];


After you get this working - you can substitute your own transformation calls that mirror the OpenGL ones so you no longer have to pull the matrix back from the graphics card if its too slow.

hth,
Codemattic

Hi, Codemattic.
For some reason, m[12]->[14] isn't correct after the glGetFloatv. It looks like the camera matrix is factored in. I'm guessing the x,y,z of the orbit should be near 40, 0, 9300. but the orbit matrix shows z at -2600. Adding in the camera matrix[14] value (-6776) gets closer to correct, but when i rotate the camera, those values go crazy.

I'm orbiting around 0,0,0.

I'm kinda doing what you have there (I leveraged some code from graphics gems that does post-concatenate of a y-axis rotation (??) - no too happy with the two-dim array - will fix later):

...
glTranslate(0,0,9380)
glRotatef(orbit, 0.0f, 1.0f, 0.0f);
glGetFloatv(GL_MODELVIEW_MATRIX, matrix);

cosTheta = cos(DEG_TO_RAD(orbit));
sinTheta = sin(DEG_TO_RAD(orbit));
for(i = 0; i < 4; i++){
t = matrix[i][0];
matrix[i][0] = t * cosTheta + matrix[i][2] * sinTheta;
matrix[i][2] = matrix[i][2] * cosTheta - t * sinTheta;
}
...
after glGetFloatv:
matrix[12] = -21 (makes sense - orbit started out x = 0, except the negative value makes me nervous)
matrix[13] = 0
matrix[14] = -2604 (should be around 9300)

after for loop:
matrix[12] = -358
matrix[13] = 0
matrix[14] = -2579

Am I on the right track or off into the weeds?

thanks,
charon

Quote:It looks like the camera matrix is factored in

did you glMatrixMode(GL_MODELVIEW); and glLoadIdentity(); before you translate and rotate?

If you provide a link to a small project which shows the problem Ill be happy to take a closer look at it.

Im going to search my code and see if I can find a better example for you.

This thread might be useful:

http://www.idevgames.com/forum/showthread.php?t=6466

codemattic Wrote:did you glMatrixMode(GL_MODELVIEW); and glLoadIdentity(); before you translate and rotate?

If you provide a link to a small project which shows the problem Ill be happy to take a closer look at it.

Im going to search my code and see if I can find a better example for you.

codemattic:

I'm not sure if I can make a small project or clean this up to be legible. I know I'm not breaking any new ground here. As I mentioned earlier - lighting, frustum culling and also collision detection. It would seem to me that finding the position of a rotated body would have been done a gazillion times. But I'm a newbie, what do i know. Thanks for the help.

Here is pretty much the front end:
====
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glMatrixMode(GL_MODELVIEW);
glLoadIdentity();

if(isViewFront == YES){
glScissor(0, 16, (GLsizei)sceneBounds.size.width, (GLsizei)sceneBounds.size.height);
glEnable(GL_SCISSOR_TEST);
}

glRotatef((float)viewAngle, 0.0, 1.0, 0.0);

// call camera rotate/xlate routine
[[self ship] handleHelm];

glGetFloatv(GL_MODELVIEW_MATRIX, camMatrix);

[[self camera] calculateViewFrustum];

glFrontFace(GL_CCW);
glEnable(GL_CULL_FACE);

glColor3f(1.0f, 1.0f, 1.0f);

glPushMatrix();
[[self starSphere] drawStarSphere];
if([[self camera] sphereInFrustum:0.0f y:0.0f z:0.0f
rad:[[self skySphere] radius]] == YES){
[[self planet] drawTexturedSphere];
[[self skySphere] drawSkySphere];
}
glPopMatrix();

glColor3f(1.0f, 1.0f, 1.0f);
glPushMatrix();
glRotatef([[self moon] rotate], 0.0f, 1.0f, 0.0f);
glTranslatef(0.0f, 0.0f, DISTANCE);
glRotatef([[self moon] orbit], 0.0f, 1.0f, 0.0f);
glGetFloatv(GL_MODELVIEW_MATRIX, moonMatrix);

// previously posted code is here

[[self moon] drawMoon];

glPopMatrix();
====

kberg Wrote:This thread might be useful:

http://www.idevgames.com/forum/showthread.php?t=6466

Well, that's certainly something to chew on. THANKS!!!

charon Wrote:I have a sphere orbiting another sphere, (say a moon orbiting a planet) using glRotate. How do I determine the xyz positon of the moon at any particular point in its orbit? I need this info for frustum culling and lighting.

Boy is my face red!! My bad. I seem to have wandered down a path that was totally unnecessary. I've figured out that if the orbit radius is the hypotenuse and the orbit angle is theta, then a little sin/cos acton produces the necessary x, z coordinates. At least I learned a tad more about matrices.

My sincere apologies to this forum and to the helpful contributors for wasting their time. I shall make a modest contribution to this site in the form of a "stupid tax."

Kindest regards,
charon

if your calculating the coordinates yourself after openGL does glRotatef your effectivly doing the same calculations twice... You might as well do the matrix calculations on your coordinates and just drop glRotatef.

If for no other reason Matrix math is essential to 3d programing and this is the perfect excuse to learn it.