Fast iterative solution of elliptic control problems in wavelet discretization (Q2499813)

The authors study the full iterative solution of partial differential equation (PDE)-constrained elliptic control problems in wavelet discretization on uniform grids using conjugate gradient (CG) schemes. They restrict themselves to the simplest class of linear-quadratic elliptic control problems with distributed control since the constraints can still be formulated weakly in terms of a single elliptic PDE. They derive a fully iterative scheme which consists of an inexact CG scheme for the control as an outer iteration in which two systems for the state and the co-state are solved approximately in interior iterations by CG schemes. In addition they use a nested iteration strategy. The authors propose a fully iterative method employing inner and outer CG schemes with optimal preconditioning which do not require the selection of such parameters to guarantee convergence. Numerical results discussing the quality of solutions, iteration numbers, and convergence histories are provided.