No label defined (Q4942489)

This paper is on an abstract model (problem \({\mathcal P})\) for boundary control problems: \[ \inf \{J(y, \pi) \mid y, \pi) \in Y \times \Pi_{\text{ad}}, G(y, \pi) = 0,\;g(y) \in {\mathcal C}\}. \] Here, \(Y\) is the state space, \(\Pi_{\text{ad}}\) is the admissible control space and \(G(y, \pi) = 0\) is the state equation \((G(y, \pi)\) a nonlinear operator). Control constraints are included in the definition of \(\Pi_{\text{ad}}\) and the last condition \(g(y) \in {\mathcal C}\) takes care of state constraints, where \(g : Y \to\) a Banach space \(Z.\) Problem \({\mathcal P}_\delta\) \((\delta > 0)\) is the same, except that the approximate version \(\inf_{z \in {\mathcal C}}\|z - g(y)\|\leq \delta\) of the state constraint is used. In the relaxed problem \(\mathcal RP,\) \(\Pi_{\text{ad}}\) is replaced by the space \({\mathcal R}_{\text{ad}}\) of relaxed controls. The relaxation is proper if the infimum for \(\mathcal P\) and \(\mathcal RP\) are the same. It is known (see references in the paper) than, under suitable assumptions, the relation is proper if and only if \(\inf {\mathcal P} = \lim_{\delta \to 0}\inf {\mathcal P_\delta},\) and proofs use continuity of the map \(r \to y_r\) from relaxed controls \(r\) to solutions \(y_r\) \(({\mathcal R}_{\text{ad}}\) is given the norm associated with the weak-star topology). When the model is applied to semilinear parabolic equations in bounded domains with control applied on the Dirichlet boundary condition, \(r \to y_r\) is no longer continuous, and the authors generalize the results on properness of the relaxation using a different approach.