Closed Loop Subspace Identification

摘要

A new three step closed loop subspace identifications algorithm based on an already existing algorithm and the Kalman filter properties is presented. The Kalman filter contains noise free states which implies that the states and innovation are uncorre-lated. The idea is that a Kalman filter found by a good subspace identification algorithm will give an output which is sufficiently uncorrelated with the noise on the output of the actual process. Using feedback from the output of the estimated Kalman filter in the closed loop system a subspace identification algorithm can be used to estimate an unbiased model.