Z+ Up, X+ Forward

Hi,

Does anyone know a fairly simple method for using rotations and/or vectors in a crazy combination to rotate an object so that the Z+ axis is always pointing up, X+ is pointing forward (toward a point I've chosen)?

If it could roll on the X axis so that it's sorta banking properly, that'd be cool too.

I'm afraid this is a bit beyond my skill with rotations at this point.

Thanks in advance.

hey. Depends a bit, could you give us some more info, is the object physical or non-physical? how is it driven? (is it a vehicle?) There's several possibilities depending on these parameters.

From what you're saying, it sounds like you want the object's x-axis to always be level with the horizon. Or maybe you actually want the object to never roll (this is how it can always be 'upright' on its z-axis). The two requirements to have sort of conflict. Anyways, to do the latter, here's some code:

// Remove 'roll' from a rotation's x-axis and return the result (Originally posted on the forums by Seifert Surface, maybe)..
rotation upright(rotation current)
{
    vector fwd = llRot2Fwd(current);
    vector left = llVecNorm(<0.0, 0.0, 1.0> % fwd);
    return llAxes2Rot(fwd, left, fwd % left);
}

It's a physical object. It can pitch, roll and yaw. I just don't want it to end up upside down. Z+ doesn't need to be exactly straight up.

ah because of the infamous gimbal lock issue. sometimes when you make large angular changes to some axis, another will jump from say, 0 to 180. I know exactly what you mean. Sadly, I don't have a solution.

Edit: I see a clue here maybe:

/54/63/146184/1.html

seems using quaternions may help ensure it doesn't happen.

trying to remember exactly... I belive you check an offset vs object rotation and then apply ether a pitch or roll... I can't think of a simplified version but....

rotation vRot90roll = llEuler2Rot( < PI_BY_TWO, 0, 0> );
vector test = <0,0,1> * llGetRot();
if (test.z < 0){
  if (test.y < 0){
    llSetRot( vRot90roll * llGetRot() );
  }else{
    llSetRot( (ZERO_ROTATION / vRot90roll) * llGetRot() );
  }
}

should be close to the right effect. or you could use PI (180) and just skip the second test set

Hewee's Rotation Tip #1 - Never, ever, EVER use Euler angles.

Seriously, think in terms of which axis you want to rotate around and by what angle, or the direction your coordinate axes (local x, y, and z axes) should be pointing. Then use functions like llAxisAngle2Rot(), llRot2Fwd(), llAxes2Rot(), etc. Even llRotBetween() can be your friend, though it's a little less clear and explicit in some cases and you have to be careful to think about what rotation axis it is going to result in.

=P thats why I converted the euler angle to a rot to give it valid context to operate in (unless you were refering to Dariens comment?)

actually, now that I think about it rot between IS the answer if you want a soft limit.

gVecUp = ;
gFltLimitUp  = 80; //--  90deg from straight up = the horizon

vector vPosDorsal;
fEnsureUp( rotation vRotCurrent ){
     //-- same as vPosDorsal = llRot2Up( vRotCurrent ); in our case, since object rots are normalized already
    vPosDorsal = <0.0, 0.0, 1.0> * vRotCurrent;

     //-- check if the dorsal position is below our limit
    if (gFltLimitUp > vPosDorsal.z){
         //-- force it up to the limit
        llSetRot( llRotBetween( vPosDorsal, <vPosDorsal.x, vPosDorsal.y, gFltLimitUP> ) * vRotCurrent );
    }
}

default{
    state_entry(){
        gFltLimitUp = llCos( gFltLimitUp * DEG_TO_RAD );
    }

    //-- the rest to the script
    //-- call function with: fEnsureUp( llGetRot() );
}

I think this would work... untested.

this will take it to the nearest limit angle, so it will usually affect the forward vector... anybody got a simpler one that will only roll?

Wow, looking at this code it looks like this problem is a lot more complicated than I thought.

If it is a vehicle, use the vertical attractor (VEHICLE_VERTICAL_ATTRACTION_EFFICIENCY and VEHICLE_VERTICAL_ATTRACTION_TIMESCALE). Otherwise, maybe try a weak llRotLookAt() that you update dynamically so it doesn't mess up the forward axis?

not so much... set a limit for the top to be facing, ceck ing the top is outside, if not move it to that limit...

getting that change to be just around the x axis (roll) seems a little harder, but I'm sure I missed something obvious and easy. it's really only 4 lines of code and some variables.