2020/05/13 by Karthik R. Ramaswamy, Ramaswamy, Karthik R., Giulio Bottegal +3
Computer Science · Engineering · #Control Systems and Identification #FOS: Electrical engineering #Fault Detection and Control Systems #Gaussian Processes and Bayesian Inference #Systems and Control (eess.SY) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2005.06266
openalex publication_date 2020/05/13 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
In order to identify one system (module) in an interconnected dynamic\nnetwork, one typically has to solve a Multi-Input-Single-Output (MISO)\nidentification problem that requires identification of all modules in the MISO\nsetup. For application of a parametric identification method this would require\nestimating a large number of parameters, as well as an appropriate model order\nselection step for a possibly large scale MISO problem, thereby increasing the\ncomputational complexity of the identification algorithm to levels that are\nbeyond feasibility. An alternative identification approach is presented\nemploying regularized kernel-based methods. Keeping a parametric model for the\nmodule of interest, we model the impulse response of the remaining modules in\nthe MISO structure as zero mean Gaussian processes (GP) with a covariance\nmatrix (kernel) given by the first-order stable spline kernel, accounting for\nthe noise model affecting the output of the target module and also for possible\ninstability of systems in the MISO setup. Using an Empirical Bayes (EB)\napproach the target module parameters are estimated through an\nExpectation-Maximization (EM) algorithm with a substantially reduced\ncomputational complexity, while avoiding extensive model structure selection.\nNumerical simulations illustrate the potentials of the introduced method in\ncomparison with the state-of-the-art techniques for local module\nidentification.\n