An intelligent gradient method and its application to parameter identification of dynamical linear processes

An intelligent gradient method is proposed for optimization problems in dynamical environments. This is an extension to the work of Zhou et al. (1993). According to a measure of dynamical performances, intelligent strategies determine recursive actions which improves dynamical performances. The proposed method is directly applied to parameter identification of dynamical linear systems in the form of intelligent recursive least squares (IRLS) algorithm. Related simulations are performed to investigate some of the properties, e.g., convergence and robustness. A significant saving in the computational cost can be achieved by the IRLS algorithm with almost no sacrifice of the estimation accuracy and of the parameter tracking ability