Economic Norm-Optimal Iterative Learning Control of a Left Ventricular Assist Device

In clinical practice left ventricular assist devices are usually driven at constant speed and thus not adaptive to changes in hemodynamic conditions. In this paper, a norm-optimal iterative learning control algorithm with an economically motivated cost function is developed, which aims at keeping ventricular volume and pressure in a predefined valid range instead of tracking specific setpoints. The algorithm is tested in silico to changes in preload and afterload. Both experiments represent naturally occurring types of parameter variations. A comparative assessment is performed between the proposed control algorithm and a constant speed controller as well as a norm-optimal iterative learning control algorithm, which is designed to shape the end-diastolic volume. The results confirm that the algorithm is able to keep ventricular volume and pressure in predefined ranges to prevent e.g. ventricular suction and dilatation. Furthermore, it offers the physician the opportunity to implement different training protocols by adapting the economic cost function.