A deployment-friendly decentralized scheduling approach for cooperative multi-agent systems in production systems

Abstract

Decentralized control paradigms are becoming more and more attractive in an ever-changing commercial environment, where there is a strong trend towards smaller production lot sizes. Whereas centralized scheduling might find a global throughput optimum (even at high computational and implementation cost), decentralized scheduling decisions in a multi-agent system are much more manageable and agents are more robust to handle any interruptions that might take place on the production floor. Compared to a centralised architecture, the development, testing and commissioning is definitely more complex, as it requires the availability of the physical units. Yet these aspects are not visited frequently by research activities.This paper details a novel implementation approach of a multi-agent based production control, that was developed for a lab-contained production environment that serves as test-bed for decentralized scheduling algorithms, with both a nominal operational mode and a simulation mode. The latter one is introduced to ease up the deployment process of the system. The description of the new approach is illustrated with different examples.