Improving the Intersample Behavior by Using Supervised Switches Among a Set of Discrete Adaptive Models and Multirate Sampling
Aitor Bilbao-Guillerna, Manuel de la Sen, Santiago Alonso-Quesada
A multiestimation adaptive control scheme for linear time-invariant (LTI) continuous-time plant with unknown parameters is presented. The set of discrete adaptive models is calculated from a different combination of the correcting gain β in a fractional order hold (FROH) and the set of gains to reconstruct the plant input under multirate sampling with fast input sampling. The reference output is given by a continuous transfer function in order to evaluate the continuous tracking error of all the possible discrete models. Then the scheme selects online the model with the best continuous tracking performance. The estimated discrete unstable zeros are avoided through an appropriate design of the multirate gains so that the reference model might be freely chosen with no zeros constrains. A least squares algorithm is used to estimate the plant parameters. However, only the active model is updated by using a least squares algorithm. The rest of possible models are updated by first calculating an estimated continuous transfer function, which results to be identical for all the models.