An Evaluation Method for Multi-Agent Systems

The growing employment of Multi-Agent Systems (MASs) in several domains of everyday life has provided the impetus for much research into new tools and methodologies for their design and implementation. But up to now, few works have focused on evaluation of these MASs, and none of these considered characteristics such as the rationality, the autonomy, the reactivity and the environment adaptability of the agents in the MAS. We believe these characteristics affect the whole performance of these systems and are connected to the complexity of the environment where the agents act. In this paper we propose an evaluation method for static multi-agent systems. The method, based on the Goal-Question-Metric approach, allows evaluation of these same MAS characteristics and combines two analysis perspectives of these systems: intra-agent and inter-agent. We also report the use of the defined approach to evaluate the GeCo_Automotive system's MAS.