Fast-array Recursive Closed-loop Subspace Model Identification

Abstract

In this paper a subspace model identification algorithm is presented that can be implemented recursively to track slowly time-varying linear systems operating in open loop and closed loop. Particular attention is paid to the computational cost and tracking performance of the developed identification algorithm. The identification problem is described by only two linear problems. The computational complexity is reduced by using array algorithms to solve these linear problems and exploiting the structure in the vectors. This results in a fast implementation of the developed recursive identification algorithm. The effectiveness of the proposed algorithm in comparison with existing methods is emphasized with a simulation study on a time-varying closed-loop system.