Adaptive backstepping control of a completely unknown permanent magnet motor

We consider adaptive backstepping (AB) control of an interior permanent magnet (IPM) motor. We propose to use artificial neural networks, or neuro-fuzzy models to approximate unknown nonlinear functions in each stage of the backstepping procedure. In this case no regression matrix need to be found and ´liner-in-the-parameter´ assumption is not necessary. The last layer coefficients of the neural network are modified on-line by the differential adaptive law. We demonstrate that adaptive backstepping technique is able to control properly a completely unknown IPM machine.