Identification of switched FIR systems with random missing outputs: A variational Bayesian approach

Identification of switched finite impulse response (FIR) systems in the presence of random missing outputs is investigated in this paper and the practical problems of unknown number of local models and unknown switching mechanism are handled. From a Bayesian perspective, the probabilistic model for describing the identification problem is constructed and the algorithm to estimate all of the unknown parameters is derived by using the variational Bayesian (VB) approach. In addition, the number of local models can be selected based on the probability of each local component, and the predicted output can be obtained as the output of the local model that takes effect. A simulated example and the mass-spring-damper system are explored to illustrate the efficacy of the developed algorithm. (C) 2020 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.

Automated summary

This paper investigates the identification of switched FIR systems in the presence of random missing outputs, addressing the practical problems of unknown number of local models and unknown switching mechanism. A probabilistic model is constructed from a Bayesian perspective, and an algorithm to estimate all unknown parameters is derived using the VB approach. Results from simulated examples and the mass-spring-damper system demonstrate the efficacy of the developed algorithm.