An experimental application of a state-space predictive controller is presented. The controller is used for low-level control of a gas conditioning unit, an integrating process with two inputs and two outputs, in a real-life environment. An original approach to the tuning of the controller's feedback action for efficient disturbance rejection performance and robustness is introduced. Feedback action is based on the Kalman filter. The disturbance model is suitable for unstable processes. The method is supported with robustness analysis using nominal sensitivity functions and a set of possible true models. The application was successful, and the load disturbance rejection performance and the robustness results were improved in comparison to the capabilities of standard control.