A Framework for Automatic Initialization of Multi-Agent Production Systems Using Semantic Web Technologies

Abstract

Mass customization and global competition require Cyber-Physical Systems of Systems (CPSoS) to become increasingly flexible. Modern CPSoS have to be able to create a wide and versatile variety of products, which takes centralized approaches to their limits. In addition, they have to produce these products as quickly as possible. Hence, they must be able to react promptly if problems arise, such as the failure of a single machine. Modern agent-based production systems provide the flexibility required to cope with these challenges. While resource agents (RAs) represent the available resources, i.e., machines, such as robots, individual customer orders can be represented by so-called product agents (PAs). However, a challenge in the design of agent-based production systems is still the amount of communication and computation that is necessary online. The PAs have to communicate their requests and the RAs their capabilities and capacities. On this basis, PAs must compute the appropriate production sequence. We propose to automatically initialize every agent with a knowledge base (KB) created a priori using semantic web technologies (SWT). On the one hand, the KBs of RAs describe the RAs' capabilities in terms of product features and production processes. Every KB of a PA, on the other hand, expresses all possible production sequences based on the customer specification and the CPSoS in question. This allows consistency checks regarding the specification as well as more purposeful communication that focuses on aspects that actually need to be determined at runtime, such as the resources' current capacities or failures. The framework presented aims to reduce both the communication and computational load necessary at runtime for agent-based CPSoS.