Online identification of linear discrete-time multivariable systems

A unified recursive algorithm is presented for identifying linear discrete-time multivariable systems from the input/output data which may be contaminated with noise. The system is represented in the form of a transfer-function matrix, and decomposed into subsystems corresponding to each row. The proposed algorithm is based on determining the order of each subsystem utilising the residual-error technique. This is followed by estimation of parameters using a recursive adaptive least-squares algorithm. Results of simulation are included.