An Investigation of Reorganization Algorithms
Christopher Zhong · 2006
Abstract Organization-based multiagent systems are designedto be highly adaptable systems. These systems areable to adapt themselves to changes that occur inthe environment. One of the ways to achieve adapt-ability is through reorganization. Currently, thereare a number of organization models that achieveadaptability with reorganization. This paper looksat how reorganization might occur for the Organi-zation Model for Adaptive Computational Systems.This paper highlights the shortcomings of the cur-rent model with respect to general-purpose reorga-nization algorithms and provides some suggestionsthat could improve efficiency of these reorganizationalgorithms. 1 Introduction As technology presses forward, more complex tasksarebeingdelegatedtocomputationalsystems. Gen-erally, these systems are distributed and expectedto adapt to changes in their environment. Whiledistributed systems offer increased reliability andaccess to distributed resources, adaptive systemscontinue to perform effectively while reacting totheir dynamically changing environments.One approach to building adaptive, distributedsystems is that of multiagent systems. However,early multiagent systems were typically designedwith a set of predefined goals and emphasized indi-vidual agents and their interactions. This resultedin adaptivity at the agent-level with system-leveladaptation being a byproduct of the agent-leveladaptation.To achieve system-level adaptation, a system-level mechanism is required. Such a mechanism isthe focus of a number of ongoing research effortsbased on an organizational metaphor. Various re-search groups are looking to provide mechanismsthat guide a group of agents by specifying high-levelobjectives within a predefined organization struc-ture. [1, 2, 3]This paper investigates the feasibility of reorga-nization algorithms for the Organization Model forAdaptive Computational Systems (OMACS), pro-vides some suggestions to improving OMACS withrespect to these algorithms, and highlights somecharacteristicsthatcanbeusedfordesigning“good”and “efficient” OMACS models.