Youla Parametrization
The Youla parametrization describes every stabilizing controller for a plant with a single free stable transfer function, making closed-loop maps affine.
All stabilizing controllers
Given a plant, the set of controllers that stabilize it looks complicated, but Youla and Kucera showed it has a clean description. Starting from any one stabilizing controller and a coprime factorization of the plant, every other stabilizing controller is generated by a single free parameter Q, where Q ranges over all stable transfer functions. Choose any stable Q and you get a stabilizing controller; every stabilizing controller arises this way.
Affine closed loops
The reason this matters is that the important closed-loop transfer functions, sensitivity and complementary sensitivity, become affine in Q. A cost that is convex in the closed-loop maps is therefore convex in Q. This converts controller design from a non-convex search over K into a convex search over the free stable parameter, which is the conceptual foundation of H2 and H-infinity synthesis.
Design use
Practically, one parametrizes the closed loop, writes the specification, minimizes over Q, and then reconstructs K. Because Q is unconstrained apart from stability, one can add convex constraints, such as interpolation conditions at plant zeros, without leaving the convex world.
- Free parameter Q ranges over all stable transfer functions
- Closed-loop maps are affine in Q
- Convex performance costs stay convex in Q
- Reconstruct the controller K from the chosen Q
A subtlety is that Q can be infinite-dimensional; practical design restricts Q to a finite basis, such as a series of stable modes, turning the optimization into a tractable finite program. This is the model-matching approach.
The parametrization also clarifies internal stability: only controllers of the Youla form avoid unstable pole-zero cancellations with the plant. It connects to loop transfer recovery and underpins the convexity that makes LMI methods applicable.