The Impact of Multi-Agent Reorganization on Efficiency

Multi-agent systems constitute a well-known paradigm to develop complex systems thanks to their ability to process dynamic situations. One of the mechanisms used to ensure dynamic responses is reorganization. It allows agents to dynamically adjust their roles and interactions in response to changes in the environment. Although reorganization improves system flexibility and fault tolerance, its impact on performance efficiency remains poorly explored. This paper presents an empirical study that investigates the effect of reorganization on MAS efficiency by defining and applying a set of customized performance metrics based on the ISO/IEC 25010 standard. The proposed metrics are evaluated based on two versions of RoboCup soccer simulation: one with reorganization mechanisms and another without. Testing the two versions across different scenarios shows that reorganization introduces significant costs, particularly in terms of memory usage (more than double), but contributes to a more balanced distribution of execution time and better collective agent behavior. These results provide valuable insights for the development of efficient multi-agent applications.