LEARNING CONTROLLER

摘要

PURPOSE:To secure the coincidence between an output vector and a target value vector by providing an output vector storage means and an output change command vector production means. CONSTITUTION:An output vector storage means 118 and an output change command vector production means 120 compare the elements corresponding to the vector storage value Yo1d storing an output vector Y(k) obtained by the k-th trial and a target value vector Yd with each other and operate the direction where the vector Y(k) should change in order to get close to the vector Yd to hold this direction at an output change command vector Yd. Then an optimum input change vector selection means 112 selects an input change vector DELTAXi with which an output change estimated vector DELTAYi is equal to the vector DELTAYd. Based on the vector DELTAXi, the (k + 1)-th trial is carried out. Thus an output vector Y(k + 1) gets close to the vector Yd compared with the precedent output vector Y(k). As a result, the coincidence is secured between both vectors Y and Yd with repetition of trials.