Optimal control problems for partial differential equations on reticulated domains. Approximation and asymptotic analysis (Q717793): Difference between revisions

This book introduces in the mathematical world of optimal control problems posed in reticulated domains. What is a reticulated domain? To have a simplified example in mind, consider the Euclidean plane covered by a uniform grid \(\mathcal{F}\) consisting of horizontal and vertical straight lines of mutual distance one: The grid is obtained by drawing all horizontal and vertical straight lines that intersect the y-axis and the x-axis, respectively, in integer points including zero. Then the plane is tiled by a periodic pattern based on the periodicity cell \([0,1)\times[0,1)\), it is reticulated. More complicated periodic structures are obtained by inserting thin or fat graphs in the periodicity cell. In civil engineering, such structures model e.g. flexible structures of gas, water, or traffic networks. As for material science, they are related to metallic, ceramic, or polymeric foams. This shows the great importance of reticulated domains for modern applications. The intersection \(\mathcal{F} \cap \Omega\) with an open domain \(\Omega \subset \mathbb{R}^2\) yields another reticulated domain. Multiplying the set \(\mathcal{F}\) by some small scaling parameter \(\varepsilon > 0\), a finer grid \(\varepsilon \mathcal{F}\) is obtained. Moreover, coming closer to practical settings, the (idealized) grid \(\mathcal{F}\) might be \(h\)-fattened by considering \(\mathcal{F}^h = \{(x,n+d): x \in \mathbb{R}, \, n \in \mathbb{Z}, \, |d| < h/2\} \cup \{(n+d,y): n \in \mathbb{Z},\, y \in \mathbb{R}, \, |d| < h/2 \}\). Then a periodic structure of thickness \(h\) is obtained that can be refined by multiplying with \(\varepsilon > 0\). We arrive at a family of open domains \(\Omega^h_\varepsilon = \varepsilon \mathcal{F}^h \cap \Omega\) depending on positive scaling parameters \(\varepsilon\) and \(h\), where \(h\) is understood as a function of \(\varepsilon\). Optimal control enters by standard elliptic, parabolic or hyperbolic optimal control problems posed in the domain \(\Omega_\varepsilon^h\). They might be cast in the abstract form \[ (P_\varepsilon) \qquad \inf I_\varepsilon(u,y), \quad (u,y) \in \Sigma_\varepsilon, \] where \ \( \Sigma_\varepsilon = \{(u,y) \in U_\varepsilon\times Y_\varepsilon \, | \, A_\varepsilon(u,y) = 0, \, u \in U_{ad,\varepsilon}, \, y \in K_\varepsilon\}. \) \ Control and state constraints are given by subsets \(U_{ad,\varepsilon} \subset U_\varepsilon\), \(K_\varepsilon \subset Y_\varepsilon\) of Banach spaces \(U_\varepsilon, Y_\varepsilon\). The operators \(A_\varepsilon\) stand for partial differential operators defined in the domains \(\Omega^{h(\varepsilon)}_\varepsilon\). The main focus of this interesting book is the limiting behavior of the optimal control problems for \(\varepsilon \to 0\). Difficult and interesting questions arise: Are the single optimal control problems \((P_\varepsilon)\) well posed? What about the behavior of their optimal solutions as \(\varepsilon \to 0\)? How to deal efficiently with the varying Banach spaces \(U_\varepsilon, Y_\varepsilon\)? If \(\varepsilon\) is very small, then the numerical solution of \((P_\varepsilon)\) might be difficult. Does there exist a ''good'' limit problem \((P_0)\) having the form of an optimal control problem? It turns out that it is not useful to define this limit problem just upon the limit components \(I_0,A_0, U_{ad,0}, K_0\). The correct answer leads to key words such as \(\Gamma\)-convergence, variational convergence, and homogenization that play a dominant role in the book. A good limit problem can be solved easier. Transporting back its optimal control to \(\Omega^{h(\varepsilon)}_\varepsilon\), a good suboptimal control for \((P_\varepsilon)\) is given. To deal with these questions, deep mathematical tools are needed that are contained in the book's Part I. For instance, elements of measure theory (including Hausdorff measures) and capacity theory, Sobolev spaces with respect to a measure, spaces of periodic functions and measures, and boundary value problems in associated spaces are introduced. Next, an introduction to the optimal control theory of PDEs is given, followed by a careful and thorough discussion of different aspects that are important for the limit analysis as \(\varepsilon \to 0\). This limit analysis forms the main line of Part I. Part II is devoted to problems with concrete PDEs in different types of reticulated domains. It is self-contained so that readers, who are familiar with the basic notions and are mainly interested in new results of research, might skip Part I. In principle, aspects of the limit analysis sketched above are discussed. The authors study the optimal control in thin periodic structures by mixed boundary controls, of ill-posed parabolic equations on thin periodic structures and on periodic singular graphs, of elliptic equations in domains with small holes, of elliptic and parabolic problems in thick multistructures by boundary controls, and the optimal control of incompressible flow in cylindrically perforated domains. By their book, the authors have paved the way for understanding main ideas of this difficult and important subject. This monograph is textbook and research exposition at the same time. It surveys known results and provides also new ones obtained by the authors. Important and deep prerequisites are collected in a short but informative way so that the reader is enabled to get easier access to the rich literature in the field. The results from optimal control theory are completely proved. Moreover, a great number of very nice and well written examples illustrate the main difficulties behind the questions and the reasons for posing them. The book provides a very good introduction into this important topic and may serve as the basis for a one semester course on optimal control in reticulated domains and for an associated seminary, where specific aspects of the theory can be discussed.