An Algorithm for Nonstochastic Identification of Controlled-Object Parameters
Z. Zh. Ametzhanova · Journal of Automation and Information Sciences · 1996
The set-theoretic (nonstochastic, guaranteed) approach [1-3, 5-7] is used widely in for solving identification problems when only the sets of the possible values of the object's unknown parameters are specified. This approach allows one to track the evolution of the multidimensional ellipsoidal estimates (approximations) that are guaranteed to contain the object's unknown parameters vector. According to the general schema for solving identification problems by the set-theoretic approach, every ellipsoidal estimate is an intersection of two sets. One of them is the result of processing all past information, and the other corresponds to the latest measurement of the object output. Investigations associated with the construction of set estimates of the unknown parameters vector and the solution of the identification problem in its nonstatic posing are continued in this paper. In contrast with existing algorithms, a new structural variant of the ellipsoidal estimates of the unknown parameters for the case in which several layers that correspond to the latest measurements of the object output intersect with the a-priori ellipsoidal set is proposed here.