Initial state learning method for iterative learning control of uncertain time-varying systems