No label defined (Q4355175)

In the paper the following optimal control problem is considered: \[ \min\{G(w)+ H(u): w\in H^1(\Omega),\;u\in L^\infty(\partial\Omega\}\tag{1} \] with state equation \[ \begin{cases} Aw= f(x) &\text{in }\Omega\\ \partial_{\nu_A}w= b(w,u) &\text{on }\partial\Omega\end{cases} \] and set of admissible controls \(U_{ad}= \{u\in L^\infty(\partial\Omega): a\leq u(x)\leq d\}\). The function \(G\) in the cost is assumed to be continuous for the uniform convergence, and the function \(H\) is taken convex and continuous for the \(L^q(\partial\Omega)\) convergence, for a suitable \(q>n-1\). The operator \(A\) is a second-order elliptic operator of the form \[ Aw=- \sum^n_{i,j=1} {\partial\over\partial x_j} \Biggl(a_{ij}(x){\partial w\over\partial x_i}\Biggr)+ a_0(x)w, \] the function \(f\) belongs to some space \(L^p(\Omega)\) with \(p>n/2\), and the mapping \(b:\mathbb{R}\times \mathbb{R}\to\mathbb{R}\) is taken Lipschitz continuous and monotone decreasing with respect to the first variable. The main goal of the paper is to prove the existence of optimal pairs for the optimal control problem (1). When the mapping \(b\) is linear with respect to the control variable, that is when \[ b(w,u)= b_1(w)u+ b_2(w), \] then a proof of the existence of optimal solutions is given. Another case in which the existence is proved is when the mapping \(b(w,u)\) is general but the cost functional does not depend on the control variable \(u\), that is \(H\equiv 0\). In the concluding remarks, the authors discuss about other particular situations that could be solved, but the question, with \(b\) and \(H\) general enough, seems to be open.