In the networked control of interconnected systems, the communication network is primarily used for the exchange of measurements amongst the control stations. Plug-and-play control extends the usage of this network towards the exchange of models with the aim to automatically design control stations at runtime. Therefore, every subsystem is equipped with a design agent that initially knows only the model of its subsystem. To design a control station by a design agent, first, a suitable model of the subsystem that interacts with other subsystems has to be set up. Second, local design conditions have to be found that guarantee the adherence of the global control aim. If the designed control station is finally plugged into the control equipment, the overall closed-loop system plays as desired. The focus of this thesis is to enable the design agent to accomplish the controller design. Therefore, three approaches are proposed which focus on the accuracy of the model that is used for the design with respect to the achievable overall closed-loop performance. The main result is a novel concept for the self-organised controller design by means of design agents. This concept is applied to achieve fault tolerance and to integrate new subsystems. The proposed methods are tested and evaluated through simulations and experiments on a thermofluid process and a multizone furnace.