Rotations (again) torques and rotational impulses.

I'm making a follower totally legally, that needs to follow its target and rotate to face the same way.

Sadly it needs physics to move right to do this.

So I know my current rot (llGetRot()) and the target object's rot (llDetectedRot()). Typically I need to be at pi/2 around z to follow with the right angle - is there a nicer way of doing that than converting to euler adding the pi/2 and converting back?

Is there a nice way of working out what rotational impulse I need to get the two to have the same rotation in a given time?

Currently I've got an ugly hack, detecting every second, moving with a tau of 0.95 then using llSetRot(), which gives jerking movement, and I suspect flicking physics on and off like that is probably worse on the sim than having it on whenever the target is moving.

Why not use llRotLookAt?

And you need to rotate it around the z axis?

llRotLookAt(llDetectedRot(0) * <0.0,0.0,1.0,1.0>

if that isn't quite it try

llRotLookAt(<0.0,0.0,1.0,1.0> * llDetectedRot(0))


I have the hardest time remembering what the order of operations do for quaternion multiplication

If you find it's rotating in the wrong direction just make the 3rd or 4th element of the rotation negative (either one just not both).

Stupidity?

I knew there was an easy way to do it, I just missed it in the list. Thanks Strife.

I find it useful to remember that if you rotate a vec, the order is:

vec * rot

So if you apply two rotations, you're doing:

(vec * rot1) * rot2 == vec * (rot1 * rot2)

...because associativity has to work or something is very wrong. So the one on the left gets applied first, the one on the right second.

Since I've opened this again can someone explain how a vector (3 X 1) times a quat (1 X 4 for ease) gives a vector (3 X 1) again? I'd expect a 3 X 4 matrix answer...

And that Strife, tried it out and it all works a treat.

I think it's like this...

Vector v;
Quaternion q;

Vector r; //Result

r = <v.x*q.x,v.y*q.y,v.z*q.z>
//Is identical to
r = r*q;

But then again, I might be wrong.

They're not "multiplying" as matrix multiplication. It's better to think of it as applying a rotation to a vector. If you start with a vector, and rotate it around some axis, you'll get another vector. I don't think you really need to get further into the guts of it to be able to do whatever you want to do.