Rotation matrices are used to describe the orientation of 3D coordinate frames in space, and to transform vectors between these coordinate frames.
Consider two 3D coordinate frames A and B that are rotated with respect to each other (Figure A.1). The orientation of B with respect to A can be described by a rotation matrix , whose columns are the unit vectors giving the directions of the rotated axes , , and of B with respect to A.
is an orthogonal matrix, meaning that its columns are both perpendicular and mutually orthogonal, so that
where is the identity matrix. The inverse of is hence equal to its transpose:
Because is orthogonal, , and because it is a rotation, (the other case, where , is not a rotation but a reflection). The 6 orthogonality constraints associated with a rotation matrix mean that in spite of having 9 numbers, the matrix only has 3 degrees of freedom.
Now, assume we have a 3D vector , and consider its coordinates with respect to both frames A and B. Where necessary, we use a preceding superscript to indicate the coordinate frame with respect to which a quantity is described, so that and and denote with respect to frames A and B, respectively. Given the definition of given above, it is fairly straightforward to show that
and, given (A.2), that
Hence in addition to describing the orientation of B with respect to A, is also a transformation matrix that maps vectors in B to vectors in A.
It is straightforward to show that
A simple rotation by an angle about one of the basic coordinate axes is known as a basic rotation. The three basic rotations about x, y, and z are:
Next, we consider transform composition. Suppose we have three coordinate frames, A, B, and C, whose orientation are related to each other by , , and (Figure A.6). If we know and , then we can determine from
This can be understood in terms of vector transforms. transforms a vector from C to B, which is equivalent to first transforming from C to A,
and then transforming from A to B:
Note also from (A.5) that can be expressed as
In addition to specifying rotation matrix components explicitly, there are numerous other ways to describe a rotation. Three of the most common are:
There are 6 variations of roll-pitch-yaw angles. The one used in ArtiSynth corresponds to older robotics texts (e.g., Paul, Spong) and consists of a roll rotation about the z axis, followed by a pitch rotation about the new y axis, followed by a yaw rotation about the new x axis. The net rotation can be expressed by the following product of basic rotations: .
An axis angle rotation parameterizes a rotation as a rotation by an angle about a specific axis . Any rotation can be represented in such a way as a consequence of Euler’s rotation theorem.
There are 6 variations of Euler angles. The one used in ArtiSynth consists of a rotation about the z axis, followed by a rotation about the new y axis, followed by a rotation about the new z axis. The net rotation can be expressed by the following product of basic rotations: .