On robustness issues in the progressive learning approach to adaptive control of high relative-degree systems

Progressive learning is an input design method for stable adaptive control of plants with high relative orders. In this paper, robustness of the progressive learning method is discussed. By applying an averaging technique, it is proved that the progressive learning method has a robustness property to unmodelled high-order dynamics in a certain frequency range. To improve the robustness to unmodelled high-order dynamics at a higher frequency range, a model augmentation technique is also presented. The key idea of the augmentation technique is that the reference model is augmented to a higher order model to cover a wider range of the bandwidth. A qualitative discussion is provided for the model augmentation. All the analyses and the arguments are verified by numerical simulation