Suboptimal control over semilinear elliptic equations with phase constraints. I: The maximum principle for minimizing sequences, the normality (Q1603048)

The paper considers standard optimal control problems for the semilinear elliptic equation \[ \text{div}[A(x)\nabla z(x)]+ a(x,z(x), u(x))= 0\quad\text{in }\Omega,\quad z\in H^1_0(\Omega),\tag{1} \] with additional pointwise constraints on the state \[ G(x, z(x))\leq q(x),\quad x\in\overline D,\quad\overline D\subset\Omega. \] The author gives necessary conditions which must satisfy minimizing approximate sequences of control \(\{u_s\}\). These conditions are of the type \[ \sup_{u\in U} \int_\Omega H(x,z_s(x),\psi_s(x), u(x)) dx- \int_\Omega H(x,z_s(x),\psi_s(x), u_s(x)) dx= \gamma_s\to 0\text{ as }s\to\infty, \] where \(U\) is the set of admissible controls, \(z_s\) is the solution of (1) with \(u= u_s\), \(H\) is a modified Pontryagin's function, \(\psi_s\) is the solution of the adjoint equation corresponding to \((u_s,z_s)\). The adjoint equation involves a measure supported on the set where \(|G(x, z_s(x))- q(x)|\leq \gamma_s\). Reviewer's remark: Almost the same results are in the paper of the same author in Control Cybern. 29, No. 2, 449-472 (2000; Zbl 0993.49020).