An optimal feedback on the basis of a mathematical model with test perturbations (Q1385944)

The paper deals with the construction of optimal feedback for an uncertain linear system. The uncertainty is represented by an unknown additive time-dependent perturbation parameter taking its values in given set \(W\). It is assumed that, at any current instant of time \(\tau \), the real state \(x^{*}(\tau)\) can be measured, moreover, the history of the process \(\{(x^{*}(t),u^{*}(t))\), \(t \in [0, \tau)\}\) is preserved. The specific feature of the problem statement is that the existence of a rule for the choice of a test perturbation \(v^{\tau }: [\tau , t^{*}] \rightarrow W\) for each pair of state and control functions defined on \([0,\tau)\) is assumed. A model predictive type controller is proposed: at any time instant \(\tau \), an optimal control problem with the chosen test perturbation is solved and the value of the optimal control at \(\tau \) is applied. Specific notions of ``realization of optimal feedback'' and ``optimal regulator'' are introduced and a numerical procedure for the calculations is proposed based on the so called defining equations of the optimal regulator.