Time optimal problems with boundary controls (Q5933799)

The authors consider the control system \[ \begin{aligned}{\partial y(t, x) \over \partial t} + Ay(t, x) + f(t, x, y(t, x)) &= 0 \quad (0 \leq t \leq T, \;x \in \Omega) \\ {\partial y(t, x) \over \partial n_A} + g(t, x, y(t, x), v(t, x)) &= 0 \quad(0 \leq t \leq T, \;x \in \Gamma)\end{aligned} \] in an \(m\)-dimensional domain \(\Omega\) with boundary \(\Gamma,\) where \(A\) is a second order uniformly elliptic partial differential operator. The control \(v(t, x)\) belongs to \(L^\sigma_{\text{loc}}(0, \infty; L^\sigma(\Gamma))\) \((\sigma > m + 1)\) and satisfies a constraint \(v(t, x) \in K(x)\) \(((t, x) \in [0, \infty) \times \Gamma).\) The problem also includes state constraints of the form \(\Phi(t, y(t, \cdot)) \in {\mathcal C},\) and consists of minimizing an integral functional of the form \[ \begin{aligned} J(y, v, T)=& \int_{(0, T) \times \Omega} F(t, x, y(t, x)) dt dx+ \int_{(0, T) \times \Gamma} G(t, x, y(t, x), v(t, x)) dt d\sigma\\ & + \int_\Omega L(T, x, y(T, x)) dx.\end{aligned} \] In this problem, \(T\) is also a control variable. Pontryagin's original version of the maximum principle for ordinary differential equations includes an additional condition on \(T\) (an equation satisfied by the Hamiltonian at the terminal time) which can be used to determine \(T.\) On the other hand, many versions of the maximum principle for partial differential equations lack this condition, one reason being the difficulties associated with the definition of Pontryagin's Hamiltonian (for instance, for hyperbolic systems). The main result in this paper is a maximum principle for the problem above including the optimality condition for the final time \(T\).