Synthesis of \(H_{\infty}\) PID controllers: A parametric approach. (Q1810578): Difference between revisions

The main result is an interesting structural property of proportional-integral-derivative (PID) controllers for scalar continuous-time linear systems. In theorem 11, it is shown that for a given proportional gain, the two-dimensional set of admissible integral and derivates gains with a constraint on the \(H_\infty\) norm of a weighted closed-loop transfer function (a typical performance measure in linear systems), lies in a union of convex sets. In order to prove the result, the authors build upon their previous work and derive variations of the Hermite-Biehler theorem for checking stability of a polynomial. One may however question the practical relevance of this result. Indeed, when resorting to \(H_\infty\) PID design, the authors propose to replace the straightforward three-dimensional gridding over the proportional, integral and derivative gains with a two-dimensional gridding over the proportional gain and an angle parameter (see step 4 of the algorithm in section 4). I am not sure this really brings a significant reduction of computational complexity, since at each step of the 2D gridding, a linear program must be solved. In contrast, at each step of the standard 3D gridding, only the \(H_\infty\) norm of a transfer function must be computed.