Efficient solution of the complex quadratic tridiagonal system for \(C^2\) PH quintic splines (Q5948207)

The problem at hand is the construction of \(C^2\) Pythagorean-hodograph (PH) quintic spline curves that interpolate a sequence of \(N+1\) points and satisfy prescribed end conditions. This leads to a tridiagonal system of \(N\) quadratic equations in \(N\) complex unknowns. Earlier approaches use a numerical continuation method to compute all \(2^{N+k}\) solutions \((-1\leq k\leq 1)\). Among this multitude of solutions there is a unique good PH spline that is free of undesired loops and extreme curvature variations. The present authors describe a more efficient method which take as a starting point ordinary \(C^2\) cubic splines; the equations for the PH curve are then solved by a Newton-Raphson iteration. The system Jacobian is tridiagonal and satisfies a global Lipschitz condition, yielding a closed-form expression of the Kantorovich condition for convergence of the iterations. Also, because of the structure of the Jacobian the iterations can be performed at \(O(n^2)\) cost. The paper contains extensive numerical tests.