zbMATH — the first resource for mathematics

Designs of learning controllers based on autoregressive representation of a linear system. (English) Zbl 0848.93027
Summary: Learning controllers that improve tracking performance through repeated trials are derived. The design is based on an autoregressive representation of a linear system. This input-output model can be interpreted in terms of an observer in state-space form. The control input is modified at every repetition as the system learns to produce a desired response, even in the presence of unknown repetitive disturbances. The coefficients of a nominal autoregressive model are first identified from input-output data. Using the identified coefficients, simple linear feedback learning controllers are designed that can correct for the errors that remain. An optimal learning gain matrix is also derived given the identified model. Numerical examples are provided to illustrate the proposed learning approach.

93B51 Design techniques (robust design, computer-aided design, etc.)
93B30 System identification
Full Text: DOI