The Complexity of Design and Test of Multi-Agent Control Systems: Lessons Learned in a Decade of Research Experience

Abstract

The design and test of multi-agent control systems is a complex and tough task. The complexity is mainly due to plenty of factors that a designer needs to define and analyze when building up and testing a multiagent control system. These factors are both due to the possible alternatives in designing the configuration of the control system itself and to the test-bed conditions wherein the control system is experimented and evaluated in its performances. This paper derives from experiences directly conducted by the authors in the last decade, during which a variety of control architectures, algorithms and manufacturing scenarios has been developed, tested and mutually compared in order to investigate the different behaviours of manufacturing shop floors under the control of multi-agent based systems. The main results of this comprehensive investigation are reported in this paper, as a contribution to the understanding of the main design variables - of the control system itself and the test-bed environment - and of the related issues and challenges to face when dealing with the development of distributed control architectures