June 2007 (Vol. 8, No. 6) p. 5
1541-4922/07/$31.00 © 2007 IEEE
Published by the IEEE Computer Society
Published by the IEEE Computer Society
Manufacturing and Control: Putting Agents to Work
PDFs Require Adobe Acrobat
Agent-Based Manufacturing and Control Systems
Massimo Paolucci and Roberto Sacile
CRC Press, 2004
Massimo Paolucci and Roberto Sacile wrote Agent-Based Manufacturing and Control Systems to fill a gap in the manufacturing and control systems literature by surveying agent-based technology research and the technology's applications to this area. They aim to explain how to make existing manufacturing information systems more agile using multiagent systems. They approach a wide variety of topics—including multiagent-systems design and implementation and manufacturing activities such as planning, scheduling, and control—and address a broad audience that includes researchers, managers, and technicians. They cohesively present relevant topics as well as a case study; they refer to this example throughout several chapters, and it helps tie the book together.
The first chapter introduces modern production systems, such as holonic and agile manufacturing systems. These systems' requirements (such as agility, flexibility, and adaptability) are Paolucci and Sacile's main motivation for applying agent technology.
Chapter 2 identifies the typical classes of manufacturing problems—planning, scheduling, and control—and introduces basic ways to approach these problems from an autonomous-agent perspective. In the second half of the chapter, Paolucci and Sacile introduce a small-enterprise case study, PS-bikes, that they refer to throughout the next three chapters.
Chapter 3 contains two parts. The first, comprising six sections, deals with the basic concepts and terminology of production planning, scheduling, and control, both from an integrated perspective and as separate activities. Having covered these prerequisites, Paolucci and Sacile summarize trends in multiagent system (MAS) applications, for both production and scheduling, and scheduling and control. They consider these integrated phases, rather than treating them separately. In the last section, they address some questions they intentionally left unanswered regarding their PS-bikes example from the previous chapter, finally resulting in an MAS architecture for planning, scheduling, and control.
The next chapter discusses applying modeling and simulation to manufacturing and multiagent-based simulation (MABS) in a general setting. In my opinion, this section's presentation is very general—maybe too general. The authors' only attempt to preserve the chapter's cohesiveness with the rest of the book is an example of applying MABS to the PS-bikes case study.
Chapter 5 describes useful techniques for designing and implementing MASs. This chapter has a software-engineering flavor. It briefly describes the FIPA (Foundation for Intelligent Physical Agents) abstract architecture and one of its major implementations, the JADE (Java Agent DEvelopment framework) platform. UML diagrams describing the structure and behavior of various agents from the PS-bikes case study spice up the presentation. The chapter also describes Passi (Process for Agent Societies Specification and Implementation), a methodology for designing and implementing MASs from specification to code generation and deployment.
Chapter 6 summarizes some successful existing applications of MAS to manufacturing systems. Paolucci and Sacile classify these applications as supply-chain oriented, production and scheduling oriented, or scheduling and control oriented. In Chapter 7, they give a bird's-eye view of trends and challenges in agent-based agile manufacturing systems. They first comment on the vision of next-generation e-manufacturing solutions. On the basis of this vision, they then discuss what must be refined and what remains to be done in agent-based manufacturing (assuming that agents are a suitable solution for the next generation of agile manufacturing systems).
I recommend Agent-Based Manufacturing and Control Systems as a broad introduction to some of the topics relating software agents to manufacturing systems. The book gives an overview of the problems in the growing but still immature field of agent systems and summarizes existing solutions and their applications to the manufacturing industry. Practitioners or theoretically minded researchers without backgrounds in manufacturing or agent systems could find useful information to guide them toward a more in-depth study.
However, the book suffers from a lack of uniformity (perhaps caused by the broad palette of subjects it covers) and its topics' heterogeneity (maybe caused by its survey or overview approach). Also, a glossary of the acronyms used in the book would have made reading the individual sections easier.
Costin Badica is a professor at the University of Craiova. Contact him at mailto:firstname.lastname@example.org