Improved feedback error learning with prefilter state variables and RLS criterion

This paper proposes an improved scheme for feedback error learning (FEL). In two-degree-of-freedom control systems in general, a prefilter is used to compensate the relative degree delay of a strictly proper plant. In conventional schemes of FEL, however, the feedforward controller has to learn parameter including the prefilter, although it is given in advance. The proposed scheme reduces this redundancy by means of the prefilter state variables as part of the feedforward signals. Furthermore, the learning law by Muramatsu et al. is generalized to the MIMO case under a recursive least square criterion.