Frequency Domain Design of a Robust Iterative Learning Control via Convex Optimization Techniques

This paper presents a robust design approach to iterative learning control using a PID controller that exploits frequency domain specifications for both robustness and performance. In a first constrained optimization stage, optimal parameters for the PID controller are determined aiming at a highest-possible integral gain for a fast compensation of step-like disturbance signals - directly in the trial in which they appear. The second design stage involves the optimal parametrization of both filters used in the ILC law, i.e., the learning filter in the form of a finite impulse response filter and the Q-filter, chosen as an infinite impulse response filter. Its goal is to meet performance specifications in a wide-enough frequency range. The overall design is performed efficiently by convex optimization techniques. Simulation results show that both robust and accurate tracking behavior is achieved after a few iterations only.