Construction of suboptimal controllers via iterative feedback tuning

This paper presents a tuning method of feedback controllers for the optimization of quadratic cost functions via Iterative Feedback Tuning (IFT), where we do not need any model of the plant to be controlled. First, we propose an IFT method in a generalized setting, where the systems are represented by Linear Fractional Transformation (LFT). Several merits of this problem setting are pointed out. Next, comparing the optimization in IFT with the iteration-based numerical computation method of the Riccati equation, we clarify some properties of the IFT method. In addition, we point out that if a certain condition is satisfied, the optimization problem in IFT becomes convex and we show how to obtain a new design parameters with I/O data only. Finally, a numerical example is given to demonstrate the usefulness of the proposed method.