Robust Recursive Identification of Time-Delay Systems With Skewed Measurement Noise

This article mainly considers the robust online identification of time-delay systems affected by skewed measurement noise. The shifted asymmetric Laplace (SAL) distribution is introduced to characterize the statistical properties of the skewed measurement noise. At the same time, the time-delay of the system is regarded as a hidden variable whose posterior distribution can be calculated using Bayes’ theorem. Consequently, the model parameters and time-delay can be robustly estimated within a recursive expectation-maximization (REM) framework, and all the unknown quantities can be updated in an online fashion to adaptively reflect the time-varying property of the underlying system. The utility of the proposed method is tested by several numerical examples, three public datasets experiments, the JT9D aero-engine model, and the hardware-in-the-loop (HIL) experiment for a turbofan engine system.