Models and Knowledge Acquisition Cycles for Multi-Agent Systems

Knowledge acquisition and modeling are important steps in developing knowledge-based systems and especially multi-agent systems. This paper presents a set of different models for knowledge acquisition and modeling for multi-agent systems. Our model set extends the KADS methodology to include distributed and cooperative aspects. Model templates are composed of generalized knowledge sources and guide knowledge acquisition. Model templates and knowledge sources are organized into knowledge libraries for reuse. Inspired by shortcomings of the model construction in KADS and VITAL, we propose a new approach including two knowledge acquisition cycles: a local and a global one. The local cycle is devoted to the step-wise refinement of generalized knowledge structures and models. The global acquisition cycle controls the development of the final model set. This paper details the local acquisition cycle.