In this diploma thesis closed loop identification is used to identify a model for restricted, unknown system behaviours. This type of systems cannot be identified by open loop identification because the system is unstable or it cannot be cut out from a superior system. At the beginning of this thesis a few cases with different identification and controller conditions are tested in a Monte-Carlo-Simulation study. Finding out which identification method is the best and how it will be used in its most effective way will be also part of this investigation. After the identification a model predictive controller (MPC) based on the evaluated model is designed. Due to the facts that a mathematical model already exists and a model predictive controller can handle restrictions in a good way, such a controller will be implemented. The combination of the two methods, closed loop identification and MPC, will be shown on a real non linear two axis helicopter model, which is indeed a perfect example, because the system is instable and restricted. Different tests with a MIMO and a SISO CLI and MPC will be performed to decide which structure should be used in the helicopter case. With the combination of MPC and closed loop identification should be demonstrated if and how far the system performance could be improved.