A contribution to the constructive theory of optimum observation of dynamic systems (Q1374964)

Problems of optimum observation are motivated by problems of guaranteed optimization of uncertain control systems. The essence of these problems is as follows. Consider the control system \(\dot x=Ax+bu,x\in R^n, u\in R,\) with an uncertain initial state (sensor) \(y=c'x+\xi, y\in R,\) on the time interval \(T=[0,t^0].\) Let there be given the class of piecewise continuous available controls \(|u(t)|\leq1 , t\in T,\) the terminal set \(X^{\ast}=\{x\in R^n:h_i'x\leq g_i, i=1, \dots,m\},\) and the criterion of quality \(J(u)=h_0'x(t^{\ast}).\) Under the assumption that additional information can only come from the sensor, it is required to securely (for all possible realizations of the initial state \(x_0\) and error \(\xi(\cdot))\) transfer the state at the time \(t^{\ast}\) to the terminal set \(X^\ast\) and ensure that the result for the worst possible criterion of quality be a maximum. A method for positionally solving the linear problem of optimum observation with breakable feedback is presented by \textit{R. Gabasov, F. M. Kirillova} and \textit{O. I. Kostyukova} [Vestn. Beloruss. Gos. Univ. Im. V. I. Lenina, I. Ser. 1992, 45-49 (1992; Zbl 0771.93078)]. In this note, positional solutions with closable feedbacks are examined.