Model predictive control of an anesthia workstation ventilation unit

Modern intensive care therapy as well as general anesthesia would not be possible, without respiratory support. Yet, unphysiological pressure levels and gas concentrations pose a serious risk to severely harm the patient. Advanced control schemes could improve the patient's safety and ensure the therapeutic success. Model predictive control (MPC) for instance allows to incorporate information about the patient at runtime through an internal model of the system, e.g. by using the lung compliance or airway resistance as model parameters. Furthermore, it can guarantee the satisfaction of constraints, which is useful, when considering physiological safety bounds. In this article we propose a two layered model-based control architecture for pressure controlled ventilation. The purpose of the lower layer is to approximately linearize the actuator dynamics, while the second layer implements a MPC controlling the pressure at the upper airways of the patient. The control architecture is implemented in an experimental setup, incorporating the ventilation unit of an anesthesia workstation. Initial results are presented, with the focus on the general feasibility of the chosen approach.