nach oben

2004 | Buch

Kapitel lesen Erstes Kapitel lesen

Multiagent Systems for Manufacturing Control

A Design Methodology

verfasst von: Dr. Stefan Bussmann, Professor Nicholas R. Jennings, Professor Michael Wooldridge

Verlag: Springer Berlin Heidelberg

Buchreihe : Springer Series in Agent Technology

Enthalten in: Professional Book Archive

Einloggen, um Zugang zu erhalten

Über dieses Buch

The ability of production companies to rapidly develop and deploy effective and efficient control systems is critical for success in the consumer-driven environment of contemporary manufacturing. This book presents a novel approach to the design of manufacturing control systems, based around the idea of agents, semiautonomous decision makers that cooperate to process goods and meet orders. This new methodology is DACS – Designing Agent-based Control Systems.

Developed at DaimlerChrysler’s research labs in Berlin, DACS is the first methodology specifically produced for the design of agent-based control systems. Beginning with a detailed overview of agent technologies, manufacturing control, and design methodologies, the book explains the DACS methodology and illustrates it by way of detailed case studies. The book will be of interest to researchers and practitioners in agent systems, manufacturing control, and software methodologies.

Inhaltsverzeichnis

Frontmatter

1. Introduction

Abstract

Software agents offer a new approach to designing and building complex distributed systems that significantly extends previous approaches like object-oriented or distributed computing (Jennings 2000; Zambonelli and Parunak 2003). Instead of modelling distributed systems as software programs exchanging data and commands, agent technology creates autonomous decision makers which communicate their preferences, negotiate subgoals and co-ordinate their intentions in order to achieve the individual or system goals (O’Hare and Jennings 1996; Weiss 1999). This decision-and interaction-based approach to computing makes it possible to build systems that can dynamically react to unforeseen events, incorporate different preferences and attitudes, exploit different capabilities of components, and adapt flexibly to changes in the environment. The ability of agents to adapt their behaviour during computation reduces the need for the designer to foresee all possible scenarios and changes the system will encounter (Jennings 2000). Moreover, an agent-oriented design is often a natural fit to the distributed nature of decision making in many application domains and thus increases the understandability and maintainability of the software system.

Stefan Bussmann, Nicholas R. Jennings, Michael Wooldridge

2. Agent-Based Production Control

Abstract

An agent-oriented design methodology for production control is found at the intersection of a new software technology, namely software agents, and an application domain, in this case production control. In order to understand the motivation for this particular design methodology, it is necessary to answer a set of questions:

What is production control? Why is it important and what is challenging about designing production control systems?

What is agent technology? And why does it help to meet the challenges in production control?

How is agent technology applied to production control? And why is it necessary to develop a design methodology for agent-based control systems?

Stefan Bussmann, Nicholas R. Jennings, Michael Wooldridge

3. Design Methodologies

Abstract

The need for methodologies in software development was already recognised in the late 1960s (Dijkstra 1968; Wirth 1971; Parnas 1972) and led to the introduction of the field of software engineering (Sommerville 1995). Since then, many methodologies have been proposed for the different phases of software development. In particular, there are a large number of methodologies for designing software systems, the most prominent probably being structured and object-oriented design methodologies. Many of these design methodologies claim to be applicable to any (software) design problem and must consequently also be applicable to the design of agent-based production control systems. This chapter will therefore review the main existing design methodologies that are potentially applicable to agent-based production control systems and will assess to what extent these methodologies are able to support adequately the design of such systems.

Stefan Bussmann, Nicholas R. Jennings, Michael Wooldridge

4. The DACS Methodology for Production Control

Abstract

The review of existing design methodologies in the previous chapter has shown that conventional design methodologies, such as for instance object-oriented or manufacturing control methodologies, are insufficient for the design of agent-based production control systems. This is because the development of agent-based systems places different requirements on the software models and methods to be used. The review also demonstrated that the agent-oriented design methodologies developed to date either lack appropriate models for modelling control decisions or are not sufficiently prescriptive to enable a control engineer without prior experience in agent technology to design an agent-based production control system. Chapter 2, on the other hand, has shown that agent technology provides indispensable features for future control systems and that a widespread industrial exploitation of this technology will require a design methodology for agent-based production control systems that can be applied by industrial engineers. To fill this gap between the state of the art and industrial needs, this chapter presents the DACS methodology for designing agent-based production control systems that is both appropriate for control design and sufficiently prescriptive for a control engineer with only minimal training and no prior experience in agent technology.

Stefan Bussmann, Nicholas R. Jennings, Michael Wooldridge

5. Evaluation of the DACS Methodology

Abstract

Chapter 4 presented the DACS methodology for designing agent-based production control systems. For this methodology, it is claimed that it is both appropriate for the domain of production control and sufficiently prescriptive for a control engineer with only minimal training in agent technology and no prior experience in developing agent-based systems to design successfully an agent-oriented production control system (cf Sect. 3.1.1). To substantiate this claim, it would be necessary to perform a large series of industrial case studies with a significant number of control engineers. However, this is clearly impossible within the scope of this book because for a realistic control problem the design process from the problem specification to the agent-based control design requires a significant investment in terms of human resources. A single case study would thus incur significant costs, a large field test prohibitively high costs.

Stefan Bussmann, Nicholas R. Jennings, Michael Wooldridge

6. Conclusion

Abstract

The goal of this book was to develop a methodology for the design of agent-based production control systems which can be successfully applied by a control engineer with only minimal training in agent technology and no prior experience in agent development. This chapter concludes the book by reviewing how this has been achieved, putting this work into context and pointing to possible future work.

Stefan Bussmann, Nicholas R. Jennings, Michael Wooldridge

Backmatter

Titel: Multiagent Systems for Manufacturing Control
verfasst von: Dr. Stefan Bussmann
Professor Nicholas R. Jennings
Professor Michael Wooldridge
Verlag: Springer Berlin Heidelberg
Electronic ISBN: 978-3-662-08872-2
Print ISBN: 978-3-642-05890-5
DOI: https://doi.org/10.1007/978-3-662-08872-2

Springer Professional

Über dieses Buch

Inhaltsverzeichnis

Frontmatter

1. Introduction

2. Agent-Based Production Control

3. Design Methodologies

4. The DACS Methodology for Production Control

5. Evaluation of the DACS Methodology

6. Conclusion

Backmatter