On the variational stability of a class of nonlinear parabolic optimal control problems (Q4361815)

The optimal control problems considered in the paper are governed by parabolic partial differential equations and can be written in the form \[ \min\int_{[0,T]\times\Omega} L(t,x,y,u,\lambda)dt dx\tag{\(\wp_\lambda\)} \] with state equation \[ \begin{cases} {\partial y\over\partial t}-\text{div}(a(x, Dy,\lambda))= f(t,x,y,\lambda)u,\\ y(0,x)= y_0(x)\text{ on }\Omega,\quad y(t,x)= 0\text{ on }[0,T]\times\partial\Omega.\end{cases} \] Here \(\Omega\) is a bounded regular domain of \(\mathbb{R}^n\), \(y(t,x)\) is the state variable, \(u(t,x)\) is the control variable, and \(\lambda(x)\) is a function which is taken in a parameter space \(\Lambda\) of functions on \(\Omega\). By using the \(\Gamma\)-convergence framework for optimal control problems developed by \textit{G. Buttazzo} and \textit{G. Dal Maso} [J. Optimization Theory Appl. 38, 385-422 (1982; Zbl 0493.49016)], the author shows, under suitable assumptions on the data, the stability of cost functionals with respect to the \(\Gamma\)-convergence, under the convergence of parameters \(\lambda\) in the space \(\Lambda\). Similarly, the stability of state equations with respect to \(G\)-convergence is shown, again under the proviso that the parameters \(\lambda\) converge in the space \(\Lambda\). As a consequence, it is shown that the multifunction \[ Q(\lambda)=\{(y, u)\text{ optimal pair for problem }(\wp_\lambda)\} \] is upper semicontinuous. The particular case of semilinear state equations is studied in Section 5 under weaker assumptions on the data.