Sufficient Conditions for Data-driven Stability of Ellipsoidal Unfalsified Control

Unfalsified control is a direct data-driven, plant-model-free controller design method, which recursively falsifies controllers that fail to meet the required performance specification, making them ineligible to actually control the plant. In this paper it is shown that sufficient conditions for stability can be derived for unfalsified control with an ellipsoidal unfalsified set, ellipsoidal unfalsified control (EUC). These conditions are: feasibility of the adaptive control problem, discarding of demonstrable destabilizing controllers and a finite number of controller switches. The latter is guaranteed by imposing a maximum volume ratio between two consecutive ellipsoidal unfalsified sets and a minimum stepsize on the controller adjustments