Design of neural network disturbance observer using RBFN for complex nonlinear systems

To solve the difficulty that the applications of disturbance observer (DOB) approaches have been limited to minimum phase systems, a neural network disturbance observer using RBFN (RBFNDOB) is proposed for complex nonlinear systems, i.e., minimum phase system and non-minimum phase system, which may be with matching or mismatching disturbances. The proposed RBFNDOB is a simple modification of the original DOB by using a RBFN to identify the inverse model of system which can track the parameter variations of real system by an on-line learning algorithm. The non-minimum phase system can be transformed into minimum phase system by constructing a pseudo-system to solve the zero dynamics in the right half plane. The RBFNDOB combining with a feedback controller can effectively suppress the disturbances of the closed-loop systems. The effectiveness and validity of the proposed control algorithm can be verified by simulations.