In optimization theories the difference (distance) between the optimal and actual case is called loss. These set of talks introduces a new theory which provides a new platform for control system optimization. This concept is based on a special decomposition of the sensitivity function of a closed-loop, defined as S=1/(1+CP), which makes possible to reformulate the classical paradigm: which is the best achievable control? Here C is the controller and P is the process.
The three decomposed parts are the design loss, the realizability loss and the modeling loss.
In the theoretical optimal case the sensitivity function is zero, which can never be reached. The decomposition helps to formulate the optimization of the contributing losses independently. The special new decomposition gives a possibility to optimize the three parts separately, but unfortunately not independently, because these components are not orthogonal. Therefore, always some iterative solution can be applied and the properties of the three parts must be investigated separately.
