System and disturbance identification for model-based learning and repetitive control (Q2702506)

The authors consider an \(n\)-th order, \(r\)-input, \(q\)-output discrete-time system NEWLINE\[NEWLINE\left\{\begin{aligned} x(k+1)&=Ax(k)+Bu(k)+B_d d(k)\\ y(k) &= Cx(k) + Du(k) + Hv(k)\\ \end{aligned}\right.NEWLINE\]NEWLINE where \(d(k)\) and \(v(k)\) are unknown periodic plant and output disturbances, respectively. One set of sufficiently rich data consisting of the excitation input \(u(k)\) and the disturbance-corrupted output \(y(k)\) is given. A formal extension of the interaction matrix formulation to the problem of system and disturbance identification is presented. With only an assumed upper bound on the order of the system and an assumed upper bound on the number of disturbance frequencies, it is shown that both the disturbance-free model and the disturbance effect can be recovered exactly from disturbance-corrupted input-output data without direct measurement of the periodic disturbances. The rich information returned by the identification can be used by a performance-oriented model-based learning or repetitive control system to eliminate unwanted periodic disturbances.NEWLINENEWLINEFor the entire collection see [Zbl 0952.00062].