In iterative learning control, several times iterative learning are needed to make the system tracking error converge, and before the system tracking error converge, tracking precision is not very high, especially in the first learning period of "inexperience". In order to solve that problem, integrating with fuzzy control, a fuzzy iterative learning control is proposed, and fuzzy control is used to compensate iterative learning control. The convergence and robustness of this fuzzy iterative learning control is proved by simulation of a double joint robot model.