This is the first chapter to consider model-based optimization as a mechanism for creating convergent Iterative Learning Control (ILC) algorithms. More precisely, Gradient-based ILC Algorithms are introduced together with a discussion of implementation and convergence properties. Robustness properties in the presence of model uncertainties are discussed and a rigorous analysis of the difficulties met in applying Gradient-based ILC to non-minimum phase (NMP) systems is characterized in terms of rapid convergence characteristics followed by NMP-induced slow convergence. Extension to point-to-point tracking task is also included and the design process is linked to classical (non-iterative) feedback design methodologies.