Multi-Rate Techniques in Non-Zero-Order Hold Identification

Abstract This paper examines the problem of parameter estimation for discrete-time, linear, time invariant stochastic models when a zero-order hold (ZOH) assumption for the input sequence is not valid. Alternatives to the ZOH assumption are investigated, including linear interpolations, higher-order hold functions and faster sampling of the input All of these solutions create a multi-rate sampling problem, which is equivalent to a system with periodically missing output data. A technique for parameter estimation in the presence of periodically missing output data is described here, and via simulation examples the performance of the algorithm is demonstrated.