Help tracking targets with child prims (rotations only)

[Fenix Eldritch]

I'm building a model ship to have rotating turrets. The idea is that when directed to "fire" on a target, the (child prim) turret would swivel around its z-axis only to face the target. Since the model will be physical and thus moving around, I need to ensure only the local z axis of the child prim changes... I've almost got it, but one last thing is tripping me up. I could use an extra set of eyes for any suggestions...

sensor(integer num_detected)
    {   //cause the child prim's forward (+X) axis to face target
        vector targetOffset = llDetectedPos(0) - llGetPos();        //target_pos - child_global_pos
        targetOffset.z = 0.0;                                       //don't care about z coordinates here
        rotation targetRot = (llRotBetween( <1.0,0.0,0.0>, llVecNorm(targetOffset) ) )*llGetRootRotation();

        llSetLinkPrimitiveParamsFast( LINK_THIS, [PRIM_ROT_LOCAL, targetRot] );
    }

 I think the problem is that last part of the calculation of targetRot.... if I don't multiply by the root rotation of the object, the child prim won't track the target accurately... it seems as the root's z rotation gets close to 180, the resulting rotation of the child turret becomes inverted.... but it still stays "upright" relative to the base.

Now if I DO multiply by the root rotation, the turret correctly tracks the target... but it no longer stays "locked" to only swiveling about its local z axis. The child prim will now always stay "upright" relative to the global frame.

(I hope I'm getting the terms right)

[Dora Gustafson]

llRotBetween is tricky in some cases
May I suggest you use llAxes2Rot which has a more firm grip on things

Something along this line:

sensor(integer num_detected)

{   //cause the child prim's forward (+X) axis to face target

    vector targetOffset = llDetectedPos(0) - llGetPos();        //target_pos - child_global_pos

    targetOffset.z = 0.0;                                       //don't care about z coordinates here

    rotation targetRot = llAxes2Rot( targetOffset, llRot2Up(llGetRot())%llVecNorm(targetOffset), llRot2Up(llGetRot()))/llGetRootRotation();


    llSetLinkPrimitiveParamsFast( LINK_THIS, [PRIM_ROT_LOCAL, targetRot] );

}

 

 

 

[Fenix Eldritch]

Thanks for your response Dora!

I spent all day playing with your code. While trying to understand how the llAxes2Rot worked,  I stumbled across your wiki page further explaining it: llAxes2Rot right and wrong and the example function calls you had Vec2RotHor() really helped me understand! From that I was able to cobble this new routuine together:

vector targetOffset = llVecNorm( llDetectedPos(0) - llGetPos() );   //normalized direction to target from turret primvector UP = llVecNorm(llRot2Up(llGetRootRotation()));               //normalized direction of "Up" for the entire shipvector LEFT = llVecNorm(UP % targetOffset);     //calculate left direction by taking cross product of up and fwd directionstargetOffset = llVecNorm(LEFT % UP);            //recalculate fwd to confine  it to the horizontal plane (relative to the ship)rotation targetRot = llAxes2Rot(targetOffset, LEFT, UP) / llGetRootRotation();  //calculate final rotation using fwd,left,up...llSetLinkPrimitiveParamsFast( LINK_THIS, [PRIM_ROT_LOCAL, targetRot] );

 And it works! No mater the orientation of the ship, the turret child prim will accurately swivel to track a target while staying confined to its local z axis. Thank you!

I just have one more question.... why must we divide the rotation by the root prim's rotation? I don't quite understand why that makes it work...

 

[Dora Gustafson]

I am glad it helped you:smileyhappy: llAxes2Rot() is strong and demanding.

---

Fenix Eldritch wrote:

I just have one more question.... why must we divide the rotation by the root prim's rotation? I don't quite understand why that makes it work...

---

It is in order to transform the global rotation into a local rotation you can use with PRIM_ROT_LOCAL.
There probably is a few more ways to do it, but if this works it is fine with me:smileywink: