Online tuning of a model-based controller by perturbation of its poles

Abstract

This study proposes an online tuning method using a model-based controller with adaptive parameters in the controller to effectively maintain the control performance and stability due to characteristic variations in the structure. Although model-based control generally provides a highly controllable performance, its performance depends on the modeling accuracy of the controlled object. Typically modeling errors, characteristics that change over time, etc. cause the performance to deteriorate. Hence, tuning of the model-based controller's characteristics is proposed as a method to adapt to the errors between a real object and its model. The main idea of the tuning method proposed in this study is that tuning the poles of the controller greatly affects control performance and stability. The tuning algorithm in the proposed method employs the simultaneous perturbation stochastic approximation (SPSA), which is well suited for optimization problems with multiple design variables. To evaluate the effectiveness of the proposed tuning method, it is applied to vibration control simulations in which the model of the controlled object is perturbed to change its physical characteristics, and then the controller is tuned to adapt to these changes. Since SPSA is a stochastic optimization method, Monte Carlo simulations are also conducted to demonstrate the effectiveness of the proposed tuning method.

Keywords

• online tuning
• simultaneous perturbation stochastic approximation (spsa)
• smart structure
• adaptive control
• self-tuning
• model-based control