Many schemes of iterative learning control (ILC) have been developed for continuous-time, non-linear dynamic systems to improve tracking performance. Two schemes, D-type and P-type, have been the bases for many ILC designs. Recently, the anticipatory ILC scheme has been introduced, on the basis of a different approach (Wang 1998a, 1999). In this paper, these basic schemes are compared from both analysis and implementation view points. The anticipatory ILC scheme is designed on the basis of a causal pair of the action taken and its resulting state variables. This approach has the anticipatory characteristics of the D-type ILC and the simplicity for implementation of P-type ILCs. A sampled-data ILC scheme is presented as another form of this anticipatory ILC scheme. Furthermore, control device saturation is taken into account and tracking error convergence results are established, with proofs. The convergence results are also provided in the presence of uncertainties, disturbances and measurement noises. Experimental results are presented to show the effectiveness of this scheme.