Least regret control, virtual control and decomposition methods (Q4950935)

The author introduces a ``least regret control'' on a simple example. Let \(A\) be a second-order elliptic operator and the state \(z\) be given by the solution of the system: \[ Az=v1_{\mathcal O},\quad \text{in } \Omega, \qquad \frac{\partial z}{\partial n_A}=g\quad \text{on} \Gamma, \] where \(1_{\mathcal O}\) is the characteristic function of \({\mathcal O}\subset \Omega\), \(v\) is the control, and \(g\) is a perturbation (or unknown) variable. \(J(v,g)\) is a quadratic functional. The solution of the following problem \[ \inf_{v} \sup_{g} \biggl\{J(v,g)-J(v_0, g)-\gamma/2 \int_{\Gamma} g^2 d\Gamma \biggr\} \] is called the ``least regret control''. By use of augmented state equations the above problem is reduced to a standard optimal control one. Moreover, the author introduces the method of virtual control, aimed at the ``decomposition of everything'' (decomposition of the domain, of the operator, etc.) and presents an algorithm. Finally, he introduces many references on the above methods and the remarks.