Using semidefinite programming for L1 identification of the ARX model

Since system identification is closely related to control theory it is quite convenient that common tools of control may prove to be useful for identification as well. Semidefinite programming is now considered as a standard tool in control theory, however its applications for identification purposes are rare. This paper shows how L₁ identification of the ARX model structure can be formulated as a semidefinite program. The way of incorporating a priori information into the identification process is demonstrated which necessitates the usage of semidefinite programming.