Problem Pivot aligning, aim constraint etc

[daniel-moreno]

Hey guys,

I am writing a script that takes two triangles (as separate objects and aligns the pivot and rotation to an edge. Kinda like this:

   _2
/_\ /
1 

So basically it was like the triangles were one shape and the faces were detached. The flow of the script is the following:

a) Gets the pivot from triangle 1
b) Gets center of each edge of triangle 2
c) Compares them to see which edge is closer
d) Places a locator on a vertex of the closes edge
e) Creates a null and aimconstrain it to the locator
f) Parents triangle 2 to this new group
g) Freezes tranforms on triangle 2

In the paper looks ok, but its kind of working 50-50. It does the right thing 50% of the times. When it goes wrong, it is slightly off about 20 degrees. and even, if I manually create a group and aim it to the locator created, still is wrong on some cases. any clues?

By the way, i even tried a non constrain solution: I discovered the euler angles comparing the edge vector to 1 0 0 (axis). and the applied the transforms to a null group with the same pivot position as the edge. STILL it gives me wrong results.

Am I doing the worng procedures? Is there another approach to it? thanks, and happy new year!

Daniel Moreno
www.danmoreno.com

[joojaa]

Do you account for the rotation order? In any case isnt it easier to use vector math?

PS: cant for earth of me understand why youd need this but hey your scrpt.

[daniel-moreno]

Well, I will explain. Im doing an animation like origami, so i need the face pivots to be aligned to an edge. So im placing a group under the face so i can just ZERO the channels the face to get to the shape i want. Yes ive tried other rotation orders, but still... what would you do in vector math? is there a way to make a group point the direction of a vector? thats what im trying to do.

[joojaa]

Well in that case use clusters istead of separate pieces.

is there a way to make a group point the direction of a vector?

yes of course the transfromation with rotation and all is just a quadruplet of vectors, because you know 3 of the vectors (along, normal and position*) you can then calculate the binormal by taking cross product along and normal. You can then just decompose your matrix, if you dont want to do so by hand then use the decomposeMatrix node (load it up with plugin manager).

You can obviously use aim too but you need to fetch up from something like the normal but hey ist the same calculation.

*along is the unit vector along edge, normal is the normal direction for example the triangle your face aligning (if its flat that is), and position is local vector to the center. Depending on the way you want it scaled oand oriented yu may need to want o fillde with the 3 first vectors order but thats about it.

[daniel-moreno]

thank you joojaa, I will try what you said. just for the record, what I was trying to do and also didnt work was to get the vector for the edge (Vertex2 - Vertex1) then I used the anglebetween -euler to get the angle difference and then applied it to the group. thats when the errors appeared 50right-50wrong. sometimes it didnt point the direction i wanted.

[joojaa]

Yes well one vector does not exactly have all the info for rotation.

[daniel-moreno]

Hey I got it working. I was missing a center pivot in the aimconstrain solution. but thanks for the info. I  will study matrix decomposition more also.

Dan