Multi-agent Collaboration in Distributed Self-Adaptive Systems

Abstract

Multi-Agent Collaboration (MAC) is a means by which agents solve complex real-world problems. It encompasses agents' coordination, negotiation and cooperation towards their shared goal. A distributed collaboration community consisting of heterogeneous agents in ever-changing environments faces varying degrees of collaborators' goals and/or polices conflicts. Therefore, the focus of my research is on collaborators' conflict management, particularly during community formation in distributed Self-Adaptive Systems. In this paper, the MAC literature is reviewed and a Multi-Agent Collaboration Community Formation (MAC-CF) taxonomy is proposed. The taxonomy adds the essential Establishment phase to the MAC process. The Establishment phase covers all the essential structural and behavioural aspects of a collaboration community formation in order to minimize agents' conflicts in the early stage of collaboration. Additionally, an architecture selection guideline will be designed to facilitate the conflict management of collaborators' goals and policies during the collaboration. The evaluation of this approach will be based on Smart-Grid (SG) scenario implementing a Multi-Agent Self-Adaptive System composed of heterogeneous agents with multiple goals and policies.