This paper proposes the adaptive parallel observer system (APOS), a multirate estimation technique for discrete time systems, to compensate for output measurements that are available only periodically and for systems whose frequency content is beyond that of the Nyquist frequency. The APOS provides the controller with stable state and system parameter estimates, which allow for the extraction of the system damping ratio and resonant frequency despite the presence of aliasing. A stability proof and error bounds are provided. The APOS is applied to the actuator dynamics of an IBM Head/Disk Assembly (HDA) System, and the APOS parameter estimates are used to extract values of the system damping ratio and aliased resonant frequency via a trained neural network.
