Subspace-based methods for the identification of multivariable dynamic errors-in-variables models

This paper analyses a multivariable errors-in-variables problem under rather general noise assumptions. Apart from the fact that both the measured input and output are corrupted by additive white noise, the output is also contaminated by a term which is caused by a white input process noise. Furthermore, these three noise processes may be correlated with each other. The solution presented here gives statistically consistent estimate of the state space matrices and it is developed in the framework of subspace model identification and is characterised by the use of instrumental variables. An example is given to demonstrate the properties of the algorithm