Designing a Simulation Middleware for FIPA Multiagent Systems
Arne Schuldt, Jan D. Gehrke, S. Werner · 2008
Multiagent systems ease the implementation of software systems to control complex processes. Instead of monolithic programs, decision-making is delegated to software agents as local entities. Like in software development in general, testing and evaluation play an important role also for multiagent systems. Particularly, because run-time interactions between agents and their effects cannot always be predicted at design time. Multiagent-based simulation is an adequate means to evaluate agents regarding their applicability in real-world operation. However, general agent development frameworks do not consider simulation-specific issues. Because they provide no means for synchronisation, an additional simulation middleware is required. Temporal criteria that are relevant for middleware design are defined in this paper. Furthermore, the actual implementation and example applications in logistics are presented.