Subdivision schemes for the fair discretization of the spherical motion group (Q955079)

The study of the rigid body displacements is a main topic in the subject of computer graphics and geometric modelling. The translational component of a rigid body displacement can be easily discretized whereas this is no longer true for its rotational component which belongs to a curved space: the spherical motion group \(SO_3\). Two totally different subdivision schemes for the fair discretization of \(SO_3\) are presented in the paper. They aim to give a discrete set of positions which are equally distributed as much as possible. Both methods are highly based on previous work developed in the last chapter (Computational kinematics) of the book by \textit{H. Pottmann} and \textit{J. Wallner} [Computational line geometry (Mathematics+Visualization, Springer, Berlin) (2001; Zbl 1006.51015)]. For instance the notion of kinematic mapping or the definition of the elliptic metric in the projective three-space can be found there. The first proposed scheme (I) is based on the 600-cell. The unit quaternion group is a double cover of \(SO_3\). Therefore, a fair subdivision scheme for discretizing \(SO_3\) can be based on those for the sphere \(S^3\subset{\mathbb R}^4\). The authors first recall the well-known icosahedral discretization of the sphere \(S^2\subset{\mathbb R}^3\) as a lower dimensional case. The four-dimensional analogue of the icosahedron is the platonic solid called the 600-cell. Since the length of the edges of the 600-cell are related to the golden ratio, then the discretization of the spherical motion group is also related to the magical constant. The subdivision scheme is based on the 600-cell as starting configuration and uses a subdivision scheme for the tetrahedra to build the next configuration. Some main facts of method (I) are: each vertex of the discretization of \(S^3\) has 12 neighbors and that its implementation and data processing needs a hierarchical data structure. In the second proposed scheme (II) the authors first outline a subdivision scheme to generate a fair discretization of the elliptic linear congruence in elliptic three-space based on the icosahedral discretization of the right unit sphere. And second, they discretize the lines of this elliptic linear congruence to obtain a fair discretization of \(E^3\) and therefore of \(SO_3\). This second step is done such that the difference between the maximal and minimal elliptic distance of neighboring grid points becomes minimal. Contrary to what happens in method (I), now each vertex of the discretization can have 12, 10 or 8 neighbors, and the implementation of the subdivision scheme makes use of a compact and efficient matrix data structure. Although both discretizations are based on different guideline principles they result in the same number of spherical motions at each subdivision step. This makes possible a comparison between them using the difference between the maximal and minimal elliptic distance of neighboring grid points. Method (II) is fairer than (I) just for the first step of subdivision. For steps 2 until 6 the grids based on the 600-cell (method (I)) are fairer than those obtained by method (II).