Algorithms for determining the model structure of a controlled system.
Miroslav Kárný · Czech digital mathematics library · 1983
MIROSLAV KARNYDetermination of the model structure of a controlled system is the necessary first step for the design of majority of sophisticated controllers including self-tuners.The Bayesian theory of system classification proved to be suitable for this task.However, the set of possible hypothesis given by orders, delays, numbers of inputs, additional measured variables including external ones etc. is too large to be fully compared.To obtain a practical tool some procedure selecting the tested hypotheses is needed.Such a procedure as well as all substantial supporting algorithms for computer-aided structure determination of a multivariate regression model is described in the present paper.