The buffered optimization methods for online transfer function identification employed on DEAP actuator

Abstract

Identification plays an important role in relation to control objects and processes as it enables the control system to be properly tuned. The identification methods described in this paper use the Stochastic Gradient Descent algorithms, which have so far been successfully presented in machine learning. The article presents the results of the Adam and AMSGrad algorithms for online estimation of the Dielectric Electroactive Polymer actuator (DEAP) parameters. This work also aims to validate the learning by batch methodology, which allows to obtain faster convergence and more reliable parameter estimation. This approach is innovative in the field of identification of control systems. The researchwas supplemented with the analysis of the variable amplitude of the input signal. The dynamics of the DEAP parameter convergence depending on the normalization process was presented. Our research has shown how to effectively identify parameters with the use of innovative optimization methods. The results presented graphically confirm that this approach can be successfully applied in the field of control systems.

Keywords

• stochastic gradient descent
• adam
• amsgrad
• deap
• system identification