Generalized Disturbance Observer Synthesis Using a Two-Stage Heuristic Algorithm

Disturbance observer (DOB) based control has been widely applied in industries due to its easy usage but powerful disturbance rejection ability. However, the existence of innate structure constraint, namely the inverse of the nominal plant, prevents its implementation on more general class of systems, such as non-minimum phase plants, MIMO systems etc.. Furthermore, additional limitations exerted on Q-filter design, i.e., unity steady state gain and low-pass nature, which narrow down its solution space largely and prevent from achieving optimal performance even if it exists. In this paper, we present a novel DOB architecture, named generalized disturbance observer (G-DOB), with the help of nontraditional use of the celebrated Youla parametrization of two degree-of-freedom controller. Rigorous analyses show that the novel G-DOB not only inherits all the merits of the conventional one, but also alleviates the limitations stated before partially. By some appropriate system manipulation, the synthesis of Q-filter has been converted to the design of reduced-order controller. Thus, a heuristic two-stage algorithm has been developed with the help of Kalman-Yakubovich-Popov (KYP) lemma: firstly design a full information controller for the augmented system and then compute a reduced-order controller. Numerical examples are presented to demonstrate the effectiveness of the proposed G-DOB structure and design algorithm.