A comparison of organization-centered and agent-centered multi-agent systems

Within the area of multi-agent systems there has recently been a development towards making the organization of such systems explicit. However, while drawbacks of classical (agent-centered) multi-agent systems have been listed, the actual advantages of making the organization explicit have not been thoroughly investigated and the approaches compared.

This paper summarizes our work with such investigation of the organization of multi-agent systems. The investigation was conducted by implementing two systems: a classical agent-centered multi-agent system (ACMAS) and an organization- centered multi-agent system (OCMAS).

The design phase of an organization-centered multi-agent system is well-defined, since there are exact requirements for specifying the organizational structure, function and obligations. Therefore, expected behavior is more easily specified in the organizational approach. It is defined using role obligations and permissions and allows the designer to specify bounds for the agents.

By taking both the agent-centered and organization-centered view on the same strategy we have gained insights about both the advantages and disadvantages of each approach. The focus has been on a single scenario, which means that not all corners of the approaches have been investigated. Even so, the results have made several differences of the approaches clear, differences that in some situations make one approach highly advantageous compared to the other.

We have chosen to consider just the Jason and Moise+ platforms and a quite complicated team-based version of Bomberman. Besides other scenarios it would be interesting to consider other organization-based models like the Agent/Group/Role (AGR) or Holonic approaches.

For full text: click here

(Author: Andreas Schmidt Jensen, Jørgen Villadsen

Published by Sciedu Press)