Organization responsible: University of Amsterdam

People involved:

• Wouter Jansweijer (project leader)
• Mattijs Ghijsen (Ph.D. student)
• Bob Wielinga (promotor)

Project description:

This research is aimed at developing a theoretical framework for a manager agent operating in a dynamic environment. This framework should enable the manager agent to adopt different coordination styles, through which a flexible organizational structure is created to deal with the dynamics in the environment.

Research in the area of organizing Multi-Agent Systems has demonstrated the need for a M.A.S. to be able to adapt its organization. When operating in dynamic environments, the ability to reorganize can be a mechanism to mitigate negative effects on the performance of a M.A.S.. In this research project we study the problem of organizational adaptation of Multi-Agent Systems (M.A.S.) operating in dynamic environments.

To be more precise, we study the decision making processes of agents that are operating in a dynamic multi-agent organization. We first provide a decision making framework that enables agents to coordinate work and to adapt their organization. The framework consists of 4 decision making processes. Two of these processes enable agents to coordination their tasks by; task decomposition (the first process) and task assignment (the second process). The framework also contains two decision processes for organizational adaptation. The first enables agents to dynamically select their coordination mechanism. The second enables them to adapt the organizational structure of the M.A.S. they are part of. The focus in this project will be mainly on the two decision making processes on organizational adaptation.

The main research questions are as follows:

• What are the required capabilities for agents to be able to dynamically adapt the organization in which they operate?
• What are the declarative and procedural components an agent should have in order to obtain these capabilities?
• What are the decision making processes agents have to be capable of in order to dynamically select their coordination mechanism? How is domain dependent and domain independent knowledge separated in these processes? What are good features for calculating the costs of coordination mechanisms?
• What are the decision making processes agents have to be capable of to adapt their organizational structure? How is domain dependent and domain independent knowledge separated in these processes? What are good heuristics for the design of organizational structures? These heuristics should lead to a more scalable approach to the design of organizations for multi-agent systems.

Publications: