This paper presents a control architecture suitable for autotrophic cultivations in a stirred tank reactor feeding a gas mixture of H-2/CO2/O-2. No metabolic information about the cultivated strain is required for the gas phase controller, which can also be used as part of the overall control for the entire process. A gain-scheduled MIMO-PI controller connected to an adaptive feedforward disturbance rejection is proposed. The latter exploits the estimated gas transfer rates of a Kalman filter solely based on a description of the non-biotic processes occurring. Thus, without further knowledge of the autotrophic strain, experiments can be run to generate data and to develop a process model including strain-specific biological details. The wild-type strain H16 of Ralstonia eutropha and two mutants, HF805 and H798, are used to demonstrate how this can be done. (C) 2020 Elsevier Ltd. All rights reserved.