A frequency-domain approach to PD-type iterative learning control