A new filter-based stochastic gradient algorithm for dual-rate ARX models

This paper proposes a new filter-based stochastic gradient algorithm for dual-rate ARX models. Algorithm analysis is based upon the Kalman filter and smoother method. The new filter applies the measurable outputs to adjust the estimated outputs during each interval of the slow sampled rate. A comparative study reveals that the present consideration makes the estimated outputs more accurate than the classical auxiliary model. A stochastic gradient algorithm is developed for the estimation of parameters using all data. The simulation made further guarantees the usefulness of the proposed algorithm.