Investigating UML- and Ontology-Based Approaches for Process Improvement in Developing Agile Multi-Agent Systems

Multi-agent systems (MAS) that link business and physical domains such as in production automation need to be reconfigured correctly and efficiently to adapt to new requirements. The standard UML-based approach only partly supports the reconfiguration process to capture agent classes and their instances, while ontologies allow modeling all aspects of MAS design in one common continuous model. In this paper we introduce a MAS development life cycle and focus on the product-specific reconfiguration of a system built mostly from reusable agents. We investigate the process variants based on (a) UML and (b) ontologies. We evaluate both process variants in a feasibility study using fundamental illustrative scenarios from an industrial production automation environment and derive lessons learned for process improvement in building sustainable MAS in the scope of production automation.