A Pattern-Based Coordination and Test Framework for Multi-Agent Simulation of Production Automation Systems

摘要

Production automation systems consist of many entities (like robots and shuttles) that interact in complex ways to provide the overall system functionality like product assembly. Multi-Agent Systems (MAS) can simulate these system entities and their interactions to better understand the system behavior according to production strategies and system configuration, which are otherwise hard to determine before their (costly) implementation. A main challenge is to coordinate the groups of agents effectively to reach a common production goal, e. g., fast and robust delivery on customer orders for assembling a set of products. In this paper, we introduce the pattern-based extension of the MAS-based manufacturing agent simulation tool (MAST) to support the simulation and evaluation of business processes in an assembly workshop: a layered coordination and test framework has been implemented and evaluated in cooperation with industry partners. Domain experts found the coordination patterns usable and useful to elicit their knowledge. The test framework allowed exploring the effects of production strategies on the system 30 times faster than with a real-world hardware workshop.