A NEW FORM OF SOLUTION OF THE LINEAR-QUADRATIC PROBLEM IN OPTIMAL-CONTROL THEORY

For dynamic objects, it is proposed to specify the desired output signals only in the class of signals that are realizable for the set of admissible controls. Under this assumption, a new form of solution, close to that of the classical theory of tracking systems, is obtained for the linear-quadratic Letov-Kalman problem. A theory of optimal control of linear stationary objects whose input signals are known in advance in a sliding time interval (objects "with vision") is developed on its basis. The frequency characteristics of optimally controlled objects of this kind are examined for the first time. Examples of possible applications are included.