In recent years the field of multi-agent systems (MAS) research has become mature. It has become apparent that the use of the MAS paradigm facilitates the modularization of complex systems in autonomous subsystems (implemented as agents) and their interactions. However, most systems that are implemented have either a very strict central control or have no central organization at all. The first type of systems use the MAS paradigm mainly to distribute the computation power. The outcomes of the system are in principle analytically determinable, but little use is made of the autonomy of the agents. The second type of systems take agents as autonomous entities pursuing their own individual goals based on their own beliefs and capabilities. In this perspective, global behavior emerges from individual interactions and therefore the final behavior of the whole system cannot be predicted, easily managed or specified externally.