Abstract
It is well known that the optimal control in a linear quadratic problem calls for feedback of all the state variables in the plant.
In many practical situations, however, not all the state variables are accessible for direct measurement and observation. Moreover, the designer may not wish to use a state reconstructor such as an observer proposed by Luenberger, because combining a state reconstructor with the plant means an increase in the performance index, the problem's complexity and the expense of the combined system. Rather, one may want to generate the control variable via a linear feedback from the output variables. In this case, one is interested in determining an optimal set of feedback gains, viz., “specific optimal control”. Such an optimal solution, in general, depends upon all the initial states.
This paper presents an approach to this class of problems where one is void of a priori knowledge concerning all or some of initial states.
