Comment by jfantl

Quaternions can be useful in robotics when you're trying to perform trajectory planning or any kind of rotation control. Quaternions provide a continuous space where every point represents a rotation, unlike rotation matrices, which exist in a much harder space to explore since most matrices do not represent pure rotations. If you have algorithms for example trying slerp between rotations, or find a path from one rotation to another under some constraint, or sample rotations near the current rotation, then the space of quaternions is a much more practical space to work in.