Quaternion Rotation Math

1 minute read

As mentioned in Understanding Quaternion Rotations, to construct a quaternion, we need to know:

  • a vector in 3-space to rotate about
  • the shortest angle to rotate about that vector

Calcualting a quaternion from two vectors

Given two vectors \(\widehat{v}\) and \(\widehat{w}\).

We can construct the vector to rotate about from \(\widehat{v}\) to \(\widehat{w}\) by taking thier cross product:

\[\widehat{r} = \widehat{v} \times \widehat{w}\]

To determine the smallest angle to rotate about \(\widehat{r}\), take the arccosine of the dot product:

\[\theta = \cos^{-1} ( \widehat{v} \cdot \widehat{w} )\]

The quaternion can then be calculated directly with \(\widehat{r}\) and \(\theta\):

\[\widehat{q} = \cos\left(\frac{\theta}{2}\right) + \sin\left(\frac{\theta}{2}\right) \frac{\widehat{r}}{\parallel\widehat{r}\parallel} \cdot \begin{bmatrix} \hat{i} \\ \hat{j} \\ \hat{k} \end{bmatrix}\]

Quaternion Conjugates

If a quaternion \(\widehat{q}_{vw}\) represents a rotation from \(v\) to \(w\), then the conjugate \(\widehat{q}_{vw}*\) represents the rotation from \(w\) to \(v\).

\[\widehat{q}_{vw} = a + b\hat{i} + c\hat{j} + d\hat{k}\] \[\widehat{q}_{wv} = \widehat{q}_{vw}* = a - b\hat{i} - c\hat{j} - d\hat{k}\]

Intuitively, the conjugate flips the vector in 3-space which we rotate about.

Rotating a vector using a quaternion

Given a vector \(\widehat{v}\) and a quaternion \(\widehat{q}\) which describes the rotation from \(\widehat{v}\) to \(\widehat{w}\):

\[\widehat{w} = \widehat{v} + 2\widehat{q}_{ijk} \times (\widehat{q}_{ijk} \times \widehat{v} + q_a \widehat{v})\]

Making a rotation matrix

Given a quaternion:

\[\widehat{q} = a + b\hat{i} + c\hat{j} + d\hat{k}\]

A rotation matrix from frame A to frame B is given by:

\[R^{A/B} = \begin{bmatrix} 1 - 2(q_{c}^{2} + q_{d}^{2}) & 2(q_{b}q_{c} + q_{a}q_{d}) & 2(q_{b}q_{d} - q_{c}q_{a}) \\ 2(q_{b}q_{c} - q_{a}q_{d}) & 1 - 2(q_{b}^{2} + q_{d}^{2}) & 2(q_{c}q_{d} + q_{b}q_{a}) \\ 2(q_{b}q_{d} + q_{c}q_{a}) & 2(q_{c}q_{d} - q_{b}q_{a}) & 1 - 2(q_{b}^{2} + q_{c}^{2}) \end{bmatrix}\]

Combining Quaternions with Quaternion Multiplication


  • a quaternion \(\widehat{q}_{1}\) which represents the rotation from space \(A \to B\)
  • a quaternion \(\widehat{q}_{2}\) which represents the rotation from space \(B \to C\)

We can multiply the quaternions to get a quaternion \(\widehat{q}_3\) which represents a rotation from space \(A \to C\):

\[\widehat{q}_{3} = \widehat{q}_{1} \widehat{q}_{2}\]

The real part of \(\widehat{q}_{3}\) is given by:

\[\widehat{q}_{3a} = q_{1a} q_{2a} - \widehat{q}_{1ijk} \cdot \widehat{q}_{2ijk}\]

The vector of \(\widehat{q}_{3}\) is given by:

\[\widehat{q}_{3ijk} = q_{2a} \widehat{q}_{1ijk} + q_{1a} \widehat{q}_{2ijk} + \widehat{q}_{1ijk} \times \widehat{q}_{2ijk}\]