A Graph Transformation Approach for Dynamic Reorganization in Multi-Agent Systems

Abstract

Nowadays, complexity and high distribution are considered as the main properties of large real-world systems such as airports and manufacturing systems. The Organization Centered Multi-Agent System (OCMAS) approach is advocated as an appropriate solution to handle the complexity and the distribution of such systems. These systems are often open, and execute under a dynamic environment with unpredictable interaction. In fact, the concept of organization that is explicitly defined in OCMAS provides a key feature to the designed system which is the “stability”. Despite that, when the Multi-agent system operates in dynamically changing environments and often unreliable communication resulting in various events originating from its external environment and/or its internal elements which de-stabilize the system state. To overcome this kind of undesirable situation, the Multi-Agent System has to reorganize its behavior and structure to continue fulfilling its overall objectives.In this dissertation, we proposed a formal approach for the reorganization of a multi-agent system statically at design time or dynamically at run-time. Indeed, we proposed to formulate the reorganization using graph transformation. In particular, we have defined a type graph to represent the multi-agent system organization and a set of rules that define the different actions that can be performed to reorganize the system. We implemented our approach using AGG. We evaluated our approach on a case study related to a manufacturing system implemented as a Multi-Agent System. The obtained results show the efficiency and the effectiveness of our approach.