Adaptive iterative learning control for nonlinear pure-feedback systems with initial state error based on fuzzy approximation.