We study an optimal control problem (OCP) subject to a PDE of elliptic type as well as state constraints. The resulting optimality system contains two PDEs, one algebraic equation and the so called complementary slackness conditions, i.e. dual products between function spaces. At this point different regularization techniques come into use.
In this paper we introduce a Barrier method as one possible way to regularize state constraints, which leads to an easily implementable path-following algorithm.
To illustrate this method, we solve first a constructed problem with known solution. Here, we can verify the rate of convergence of the path-following method. Second, a simplified hyperthermia problem in 3D is solved by using COMSOL Multiphysics.
We show how time-dependent optimal control for partial differential equations can be realized in a modern high-level modeling and simulation package. We summarize the general formulation for distributed and boundary control for initial-boundary value problems for parabolic PDEs and derive the optimality system including the adjoint equation. The main difficulty therein is that the latter has to be integrated backwards in time. This implies that complicated implementation effort is necessary to couple state and adjoint equations to compute an optimal solution. Furthermore a large amount of computational effort or storage is required to provide the needed information (i.e the trajectories) of the state and adjoint variables. We show how this can be realized in the modeling and simulation package COMSOL MULTIPHYSICS, taking advantage of built-in discretization, solver and post-processing technologies and thus minimizing the implementation effort. We present two strategies: The treatment of the coupled optimality system in the space-time cylinder, and the iterative approach by sequentially solving state and adjoint system and updating the controls. Numerical examples show the elegance of the implementation and the efficiency of the two strategies.
Solving Time-Dependent Optimal Control Problems in Comsol Multiphyiscs ba Space-Time Discretizations
(2009)
We use COMSOL Multiphysics to solve time-dependent optimal control problems for par-
tial differential equations whose optimality conditions can be formulated as a PDE. For a
class of linear-quadratic model problems we summarize known analytic results on existence
of solutions and first order optimality conditions that exhibit the typical feature of time-dependent control problems, namely the fact that a part of the optimality system has to be
integrated backward in time. We present a strategy that is based on the treatment of the
coupled optimality system in the space-time cylinder. A brief motivation of this approach is
given by showing that the optimality system is elliptic in some sence. Numerical examples
show advantages and limits of the usage of COMSOL Multiphysics and of our approach.
In this paper we present a strategy to solve parabolic optimal control problems
using available specialized elliptic PDE solvers. We aim at an indirect solution
approach, i.e. developing optimality conditions in function spaces that are then
discretized and solved. Classes of problems where optimality conditions can be derived
as coupled systems of parabolic partial differential equations are considered.
We consider a simultaneous space-time discretization. We verify that for our model
problems the parabolic forward-backward system of PDEs can equivalently be expressed
by a single elliptic boundary value problem in the space-time domain. This
fact has been used as a motivation for space-time-multigrid solution approaches,
which may also be an option in our context.
The theoretical base developed for the example problems then allows to apply
specialized elliptic PDE solvers to the optimality system without much implementational
effort. Numerical experiments for some example problems are conducted and
underline the applicability of this approach.
In the first part of this article, we have shown how time-dependent optimal control for partial
differential equations can be realized in a modern high-level modeling and simulation package. In this second part we extend our approach to (state) constrained problems. "Pure" state constraints in a function space
setting lead to non-regular Lagrange multipliers (if they exist), i.e. the Lagrange multipliers are in general Borel
measures. This will be overcome by different regularization techniques.
To implement inequality constraints, active set methods and interior point methods (or barrier methods) are widely in use. We show how these techniques can be realized in the modeling and simulation package Comsol
Multiphysics.
In contrast to the first part, only the one-shot-approach based on space-time elements is considered. We implemented a projection method based on active sets as well as a barrier method and compare these methods
by a specialized PDE optimization program, and a program that optimizes the discrete version of the given problem.