Generalization of ILC for fixed order reference trajectories using interpolation

The increasing demands for motion accuracy in high–precision mechatronics call for intelligent solutions to feedforward controller design. Iterative learning control (ILC) produces data–driven feedforward signals that give high accuracy for repeating references. However, the ILC feedforward input requires time consuming re–learning for each variation of the reference. In order to circumvent the re–learning process, this paper presents a feedforward controller design that can handle fixed order references. First, we assume that ILC is used to obtain feedforward signals for a finite number of repeating references, and that these references can be split into sections that admit a polynomial parameterization. Then, we show that a new feedforward input can be calculated from the existing ILC signals for any polynomial reference spanned by parameter–wise linear combinations of the learned references. Effectiveness of the method is shown in simulation of a coreless linear motor.