Repetitive space-learning controls are designed for current-fed uncertain permanent magnet step motors with non-sinusoidal flux distribution (the family of permanent magnet synchronous motors with cogging torque is allowed as a special case). Either semi-global rotor speed tracking is asymptotically achieved or local rotor position tracking is asymptotically guaranteed without requiring the time-periodicity of the corresponding reference signals. Simulation results illustrate the effectiveness of the presented approach in a typical electric drive control scenario, even in the presence of stator current dynamics which have been neglected at the design stage. (C) 2016 Elsevier Ltd. All rights reserved.