Abstract:
The paper is concerned with extreme distributions of maximal likelihood estimates for parameters of a closed-loop system. Expressions are given for an asymptotically normalized information matrix and a numerical algorithm for its computation is proposed. A relation is found for degeneracy of an information matrix and structural nonidentifiability of the plant in a closed-loop system. The conditions for the plant structural nonidentifiability conditions to be crude are proven in a closed-loop system where noise is introduced into the feedback. A numerical procedure is developed for design of feedback and the input signal for identification of the plant in a closed-loop system.