Solving polynomial systems using no-root elimination blending schemes

Searching for the roots of (piecewise) polynomial systems of equations is a crucial problem in computer-aided design (CAD), and an efficient solution is in strong demand. Subdivision solvers are frequently used to achieve this goal; however, the subdivision process is expensive, and a vast number of subdivisions is to be expected, especially for higher-dimensional systems. Two blending schemes that efficiently reveal domains that cannot contribute by any root, and therefore significantly reduce the number of subdivisions, are proposed. Using a simple linear blend of functions of the given polynomial system, a function is sought after to be no-root contributing, with all control points of its Bernstein–Bézier representation of the same sign. If such a function exists, the domain is purged away from the subdivision process. The applicability is demonstrated on several CAD benchmark problems, namely surface–surface–surface intersection (SSSI) and surface–curve intersection (SCI) problems, computation of the Hausdorff distance of two planar curves, or some kinematic-inspired tasks.

► An improvement of the sign exclusion test is introduced for solving well-constrained piecewise polynomial systems. ► Two linear blending schemes that detect the domains that cannot contribute by any root are introduced. ► The applicability of the method is demonstrated on several benchmark CAD problems.