Convergence performance oriented data-driven tuning method for parameterised controller design with cases investigation

This study develops a data-driven controller tuning scheme to iteratively achieve the desired objective criterion with significant improvement of the convergence performance. Differing from model-based approaches, the controller is in a linear parameterisation form and the controller parameters are tuned by fully using available collection data to address the practical difficulty in obtaining accurate dynamic model. Specifically, the internal iterative behaviour between the current parameter and the optimal parameter is firstly analysed with mathematic expression. Moreover, a novel iterative law based on the behaviour is proposed, which has the ability to directly seek the global optimal parameter that minimises the objective criterion. Furthermore, an unbiased gradient approximation based on the Toeplitz matrix is developed to simplify the practical implementation. Case studies are conducted by both simulation and experiment, and the results consistently indicate that the proposed tuning scheme not only guarantees the parameter converging to global minimum, but also possesses an excellent convergence rate. The proposed strategy essentially provides a novel data-driven controller tuning method and could be applied to practical applications.