Adaptive Learning-Based Quantized Parameter Identification of Nonlinear Sandwich-Like Systems With Memory Nonlinearity

Although several quantized parameter identification methods exist, new applications require high efficiency, novel frameworks, and adequate estimation performance. In this study, we develop a quantized parameter estimation method for sandwich-like systems with backlash nonlinearity using data-driven self-error learning. First, the quantized system is transformed into a compact identification model based on a backlash parameterized form. The estimation error information is then extracted using filtered variables via an error-feedback filter. Second, a corresponding compensation error variable is established through the eigenvalue decomposition principle to address the unfavorable effect of the regression vector on the estimator. Finally, a novel performance function is designed and used to derive an adaptive learning law using an optimization technique. Statistical assessments from numerical simulations and real-life systems further confirm the effectiveness of the proposed estimation method.