A management architecture for multi-agent systems

Agents do have certain properties, such as autonomy, learning or mobility, which are useful to design and implement complex distributed applications. A multi-agent system is a network of cooperating agents each covering a well-defined and restricted part of the solution. Although agents are in principle autonomous they have to be monitored and controlled which is the task of the management of the multi-agent system. With the example of a Virtual Project Office a concrete multi-agent system and its management issues are illustrated. We introduce a management architecture which starts from existing and widely accepted concepts from the area of integrated network and systems management. One innovative aspect of the management architecture is given by the decision to model the management components of the architecture as non-mobile agents. This results in a simple and homogeneous integration of management functionality into the multi-agent system.