A model-data integrated iterative learning controller for flexible tracking with application to a piezo nanopositioner

In precise motion systems, feedforward controller is a key component for significant performance enhancement. However, traditional iterative learning control (ILC) works efficiently under strictly repetitive reference input, and the performance of model-based feedforward controllers is limited by the non-minimum phase zeros and modeling uncertainties during executing tasks. In this paper, a model-data integrated ILC is proposed for flexible tracking, where the stable part of the identified model is utilized to compose the model-based part, and the modeling error and gain mismatch are compensated by the data-driven approach via constructing a parameterized finite impulse response filter. In order to diminish the effect of noise, an instrumental variable method is adopted in the cost criterion. The proposed controller has an analytic solution and retains stability during iterations, which is verified on a piezo nanopositioner. Comparative experimental results indicate that the proposed controller can realize flexible tracking in comparison with norm optimal ILC, and achieve the best performance compared with zero-phase-error tracking controller and polynomial basis functions feedforward controller.