Introducing a Model-based Learning Control Software for Nonlinear Systems: RoFaLT

Abstract

This paper introduces ROFALT, an open-source, model-based iterative learning control (ILC) tool for nonlinear systems, that aims at closing the gap between the theory of nonlinear ILC and successful applications. Providing a simple yet powerful syntax in MATLAB, ROFALT supports all phases of the design of a nonlinear ILC—from modeling, tuning and execution, to analysis. The theoretical basis is an optimization-based two-step approach that allows an easy trade-off between fast convergence and robustness for generic nonlinear systems. To demonstrate the efficiency of the developed tool, a simulation study on an overhead crane is performed, where a model with introduced parameter deviations is used to iteratively learn the open-loop control inputs. Special attention is paid to the comparison of different possible ways to realize the learning effect and the resulting performance. Moreover, the simplicity of considering constraints and their compliance is demonstrated, while fast convergence is observed.