Constrained data-driven model-free ILC-based reference input tuning algorithm

Abstract

This paper proposes a data-driven Iterative Reference Input Tuning (IRIT) algorithm that solves a reference trajectory tracking problem viewed as an optimization problem subjected to control signal saturation constraints and to control signal rate constraints. The IRIT algorithm incorporates an experiment-based stochastic search algorithm formulated in an Iterative Learning Control (ILC) framework in order to combine the advantages of model-free data-driven control and of ILC. The reference input vector’s dimensionality is reduced by a linear parameterization. Two neural networks (NNs) trained in an ILC framework are employed to ensure a small number of experiments in the gradient estimation. The IRIT algorithm is validated by two case studies concerning the position control of a nonlinear aerodynamic system. The results prove that the IRIT algorithm offers the significant control system performance improvement by few iterations and experiments conducted on the real-world process. The paper successfully merges the use of ILC in both model-free reference input tuning and NN training.