A closed-loop system (AUTOPILOT-BT) for the control of mechanical ventilation was designed to: 1) autonomously achieve goals specified by the clinician, 2) optimize the ventilator settings with respect to the underlying disease and 3) automatically adapt to the individual properties and specific disease status of the patient. The current realization focuses on arterial oxygen saturation (SpO(2)), end-tidal CO(2) pressure (P(et)CO(2)), and positive end-expiratory pressure (PEEP) maximizing respiratory system compliance (C(rs)). The "AUTOPILOT-BT" incorporates two different knowledge sources: a fuzzy logic control reflecting expert knowledge and a mathematical model based system that provides individualized patient specific information. A first evaluation test with respect to desired end-tidal-CO(2)-level was accomplished using an experimental setup to simulate three different metabolic CO(2) production rates by means of a physical lung simulator. The outcome of ventilator settings made by the "AUTOPILOT-BT" system was compared to those produced by clinicians. The model based control system proved to be superior to the clinicians as well as to a pure fuzzy logic based control with respect to precision and required settling time into the optimal ventilation state.