Direct Data-Driven Tuning of Look-Up Tables for Feedback Control Systems

In industry, a feedback controller with a look-up table (LUT) is often used for nonlinear systems. Although this structure is easy to understand, tuning the LUT parameters is time-consuming due to the huge number of parameters. This paper presents a direct data-driven design method for a gain-scheduled feedback controller with a LUT to directly tune the LUT parameters from single-experiment data without a system model. Specifically, conventional virtual reference feedback tuning (VRFT), which is a data-driven method, is extended and the
${L} _{2}$
norm for adjacent LUT parameters is added to the VRFT cost function to avoid overlearning. The optimized parameters are analytically obtained by a generalized ridge regression. A simulation of a nonlinear system demonstrates that the proposed method can directly obtain the LUT parameters without knowledge of the controlled object’s characteristics.