This paper proposes a computationally efficient iterative learning control (ILC) approach termed non-lifted norm optimal ILC (N-NOILC). The objective is to remove the computational complexity issues of previous 2-norm optimal ILC approaches, which are based on lifted system techniques, while retaining the iteration domain convergence properties. The computational complexity needed to implement the proposed method scales linearly with the trial length. Therefore, the approach can be implemented on controlled processes having long trial durations and high sampling rates. Robustness is accomplished by adding a penalty term on the control input in the cost function. Simulations are presented to verify and validate the features of the proposed method. (C) 2013 Elsevier Ltd. All rights reserved.