Elimination in generically rigid 3D geometric constraint systems

Modern geometric constraint solvers use combinatorial graph algorithms to recursively decompose the system of polynomial constraint equations into generically rigid subsystems and then solve the overall system by solving subsystems, from the leave nodes up, to be able to access any and all solutions. Since the overall algebraic complexity of the solution task is dominated by the size of the largest subsystem, such graph algorithms attempt to minimize the fan-in at each recombination stage. Recently, we found that, especially for 3D geometric constraint systems, a further graph-theoretic optimization of each rigid subsystem is both possible, and often necessary to solve wellconstrained systems: a minimum spanning tree characterizes what partial eliminations should be performed before a generic algebraic or numeric solver is called. The weights and therefore the elimination hierarchy defined by this minimum spanning tree computation depend crucially on the representation of the constraints. This paper presents a simple representation that turns many previously untractable systems into easy exercises. We trace a solution family for varying constraint data.