On Reduced Structure Control and the 1-norm regularisation

This paper studies the controller structure selection problem for large multivariable systems. Current approaches to this problem either rely on the control engineer´s experience and intuition, or solutions of very difficult and non-convex combinatorial optimisation problems. In this paper new computationally inexpensive analysis tools are presented which provide the control engineer with a complete set of optimal controller structures for a given system. Furthermore, the proposed tools can be used directly to assess the local structural sensitivity of the problem and identify key and important connections within the controller. These tools are applied here to the model of a real-life gas-turbine engine.