Abstract
The filtered CARMA model provides a general framework for designing self-tuning controllers based on a time-series representation. Three different state-space formulations for the filtered CARMA models are discussed. The innovations model has been commonly used but is found to be inadequate for deriving the classical self-tuners given by Åström and Wittenmark 1973; the noise-free-measurement model overcomes this inadequacy by rearranging the disturbance terms for the state-space model. Finally, the Kalman-fitter-type model provides an alternative and more general framework for deriving classical polynomial self-tuners, in addition to being used as a basis for other types of controller design.