Finite Element Methods for PDE-Constrained Optimal Control Problems
Abstract
In this course, we present theoretical and practical aspects of finite element methods applied to
discretizations of PDE-constrained optimization problems.
In the theoretical part of the course, we discuss different discretization concepts carefully taking into
account the difference between discretization of a single equation and the discretization of an
optimization problem. We derive a priori error estimates for different situations including problems
with inequality constraints as well as problems governed by semilinear PDEs. Moreover we discuss a
posteriori error estimates and adaptive mesh refinement algorithms for PDE-constrained
optimization. Last but not least, we give an overview of current research topics in this area.
In the practical part of the course the students get to know the software packages
Gascoigne and
RoDoBo (both written in C++). The finite element library Gascoigne is a flexible solver for elliptic and
parabolic PDEs and RoDoBo is a package for solving PDE-constrained optimization problems with an
interface to Gascoigne. The students learn to use these packages for simulation and optimization of
PDEs and to carry out numerical experiments. In particular, the following aspects of numerics for
PDEs will be investigated: Solving linear and nonlinear systems of stationary and nonstationary PDEs
with different types of boundary conditions, evaluation of discretization errors and functionals, mesh
generation and adaptation, and evaluation of a posteriori error estimators. Finally, the mentioned
aspects will be extended to the treatment of PDE-constrained optimization and parameter
estimation problems.
Date and Place