Consideration of disturbance response and stability margin for V-Tiger optimizing virtual time responses derived from frequency data

Abstract

Generally, in control gain tuning, settling time and overshoot are altered to be smaller through time-consuming and labor-intensive experiments. Such experiments can be automatically performed in the computer in a short time if the time response can be predicted. Kaneko et al. proposed a new approach for predicting the response of a closed-loop system using one-shot experimental data, but this approach takes into consideration the structure of the closed-loop transfer function. Thus, two methods have been proposed that do not require the structure of the transfer function. One method is based on the finite impulse response estimated from one-shot experimental data by Matsui et al., but the design guideline for an unstable system is not shown. The other method is Virtual Time-response based Iterative Gain Evaluation and Redesign (V-Tiger), which is based on the frequency components of one-shot experimental data by Kosaka et al. However, this proposed V-Tiger has the following issues: 1) Inability to predict disturbance response and input disturbance response. 2) The initial value used for optimization must be a controller that stabilizes the system. Hence, in this paper, we describe the countermeasures against these issues.