Generating and Supporting Dynamic Heterogeneous MAS

Abstract

One of the grand challenges in agent technology research is quasi-automatic development of heterogeneous and dynamic multiagent systems using agent oriented software engineering (AOSE) tools. This paper presents an innovative approach to developing and supporting multiagent systems at run-time. Multiagent systems (MASs) can and should be generated dynamically based on high-level user specifications which are transformed into a mission. Dynamically generating agents could also be offered as a pervasive service. Heterogeneity of MASs refers to diverse functionality and constituency of the system which include mobile as well as host associated software agents. This paper proposes and demonstrates on-demand and just-in-time agent generation approach which is combined with run-time support for MASs. Run-time support is based on mission cost-efficiency and objectives which enable termination, generation, injection and replacement of software agents as the mission evolves. We present the formal underpinning of our approach and describe the prototype tool - called eHermes, which has been implemented using specific agent platforms. Lessons and results of testing eHermes are reported and analyzed