A Generalized Design Method for Learning-Based Disturbance Observer

This paper presents a generalized disturbance observer (DOB) design framework that is applicable to both multi-input-multi-output (MIMO) and non-minimum phase systems. The design framework removes conventional DOB's structure constraint, which allows minimizing the H-infinity norm of the dynamics from disturbance to its estimation error over a larger feasible set. The design procedure does not require explicit plant inverse, which is usually challenging to obtain for MIMO or non-minimum phase systems. Furthermore, the generalized DOB is augmented by a learning scheme, which is motivated by iterative learning control, to further enhance disturbance estimation and suppression. Both numerical and experimental studies are performed to validate the proposed learning-based DOB design framework.