Linear differential games in a Banach space (Q1280890)

The authors have extended some results of Pshenichnyj's theory of \(\varepsilon\)-strategies in differential games from the finite-dimensional to Banach spaces. The investigations are based on applications of the Hille-Yosida theorem to provide representation formulas for solutions of the controlled system \[ \dot z = Az + u(t) - v(t); \] \[ z(0) -z^0 =\phi (x); \] \[ z^0\in L^2 [0,1] = H; \;z(t) = w(x,t); \] \[ u(t) =\widetilde u(t)\delta(x); \;v(t) = \widetilde v(t)\delta(x); \] \(z(\cdot), u(\cdot), v(\cdot) \in L^2 (0,\theta, H)\), the derivative is understood in the generalized sense. The operator \(A\) associates to every element \(z\) from \(H\) its second derivative subject to the boundary condition \[ \frac{\partial}{\partial x}w(0,t) = \frac{\partial}{\partial x}w(l,t) = 0 \] The terminal set is defined by the expression \[ M=\left\{ f(\cdot)\in H: \;c\leq\frac{1}{2\varepsilon} \int^{1/2-\varepsilon}_{1/2+\varepsilon} f(x)dx \leq d \right\}. \] The goal of the control \(u\) is to satisfy the condition \(z(\theta)\in M\), the control \(v\) has the opposite aim. Previous publications have shown that the programmed absorbtion operator \(P^{\theta}\) and the operator \(\widetilde P^{\theta}\) based on \(\varepsilon\)-strategies are identical in certain cases \((A=0\) and convex \(M\); \(M\) is \(H\)-convex, where \(H\) is the set of eigenvectors of the matrix \(A)\). In this case the operators define the set of solvability for the \(u\)-control problem. In the paper, an example for the controllable heat equation is exposed where the operators \(P^{\theta}\) and \(\widetilde P^{\theta}\) are identical for an essentially nontrivial matrix \(A=\)diag\((\lambda^1, \dots, \lambda^{n}, \dots)\), where \(\lambda^{n} \to \infty\) as \(n\to\infty\).