Internal model-based controller design using measured costs and gradients

We consider the problem of tuning a closed-loop controller for a continuous-time, linear, lumped, exponentially stable plant using measurements. With the aim of reducing the amount of computation and the effect of modelling error, a two-step tuning method is proposed based on an internal model controller structure. The first step solves a controller design problem formulated as a sequence of convex optimization problems in which the cost, constraint functions and their gradients are evaluated directly from plant measurements. The designer comfortably specifies the desired performance by adjusting the weights of a cost function composed of time and frequency domain performance functionals. Using plant frequency response data, the second step solves a convex optimization problem for a plant model required to implement the controller. Model error bounds ensure closed-loop stability. An example illustrates the design procedure.