A new observer-based approach to disturbance rejection and control in linear continuous systems is introduced. A recently developed technique of continuous deadbeat observation is exploited to estimate the process states and disturbance signal described by a known dynamic model. A realization condition in the form of linear algebraic equations is derived. When there is no reliable disturbance model available, it is suggested that a neural network-based structure be used. An application example illustrates the feasibility of the approach taken
Godkänd; 1995; 20080602 (cira)