A direct tuning of disturbance observer from experimental data evaluating noise rejection

A disturbance observer (DOB) is known as an effective control method for perturbations of a plant and/or disturbances. The DOB consists of an inverse system of the plant model and a disturbance estimation filter, which estimates perturbations and/or disturbances and compensates them. The bandwidth of compensation by the DOB depends on the gain of the filter, and a larger bandwidth is desired. On the other hand, a high gain filter causes an excessive amplification of the measurement noise in high frequencies. There is a tradeoff between compensation and noise rejection for the DOB. In this paper, we propose the design method of the DOB from input/output data considering the trade-off between compensation and noise rejection. The proposed design method optimizes tracking performance imposing two constraints; the stability constraint of the closed-loop system and the noise rejection constraint. The stability constraint is based on the Nyquist stability criterion and described by the collected data set. The constraint for noise rejection is evaluated by spectral analysis.