A randomized algorithm for simultaneous identification of linear systems

Abstract

The authors propose, for the first time in system theory, a robust version of the well-known system identification problem. The problem, which they call the simultaneous identification problem, arises in trying to identify the parameters of a time-varying system. In the past, an identifier for a time-varying system was constructed by an immediate generalization of the ideas pertaining to a time invariant system. Unless the time variation of the plant is assumed to be slow, such an identification scheme would result in poor performance. As an alternative, a probabilistic randomized algorithm is presented, and its convergence is investigated.<>