nach oben

2006 | Buch

Multiagent Engineering

Theory and Applications in Enterprises

herausgegeben von: Professor Dr. Stefan Kirn, Professor Dr. Otthein Herzog, Professor Dr. Peter Lockemann, Professor Dr. Otto Spaniol

Verlag: Springer Berlin Heidelberg

Buchreihe : International Handbooks on Information Systems

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

1 Multiagent Engineering: A New Software Construction Paradigm Multiagent systems have a long academic tradition. They have their roots in distributed problem solving in Artificial Intelligence (AI) from where they emerged in the mid-eighties as a distinctive discipline. Research in multiagent systems owes much to the work of Rosenschein on rationality and autonomy of intelligent agents, the European MAAMAW workshop series, and last but not least the famous readings of Bond & Gasser (1988) and Jacques Ferber´s book on multiagent systems (1991). It gained further by a public discussion via the Distributed AI mailing list in summer 1991, when the pioneers of the field compared in much detail the concepts of distributed problem solvers to multiagent systems. Within only five years, a new exciting field of research had been established. Now, 15 years later, the field has matured to a degree that allows the - sults of academic research to be passed on to practical use and commercial exploitation. This potential coincides with a need for much larger flexib- ity of our IT infrastructure in light of its highly distributed character and extreme complexity, but also the global character of the business processes and the large number of business partners due to outsourcing and specia- zation. Many experts claim that multiagent systems are the right software technology for the needed IT infrastructure at the right time. The appeal has much to do with the broad perspectives of multiagent systems research.

Inhaltsverzeichnis

Frontmatter

Management Summary

Peter C. Lockemann, Stefan Kirn, Otthein Herzog

What Agents Are and What They Are Good For

Frontmatter

1. Agents

Abstract

Agents are many things to many people. This is not a sound basis for engineering software agents or multiagent systems. Since this book deals with engineering such systems, the first thing to do is to agree on the notion of software agent that is to be used throughout the book. The approach will be in terms of the qualitative properties that are deemed essential or at least important.

Peter C. Lockemann

2. From Agents to Multiagent Systems

Abstract

In the previous chapter agents and their properties have been introduced. In real-world business applications, it is assumed that the benefit of agent technology is reached by dynamic interaction of autonomous agents. This interaction and co-operation forms a multiagent system (“MAS”). The organization of agents within such systems is strongly related to organization theory. The specific flexibility of MAS arises from the ability to follow predefined structures or evolve structures from dynamic interaction. In this chapter, fundamental concepts and properties of MAS are introduced with special focus on interaction and communication, roles, and structures.

Ingo J. Timm, Thorsten Scholz, Otthein Herzog, Karl-Heinz Krempels, Otto Spaniol

3. Flexibility of Multiagent Systems

Abstract

The most important promise of multiagent technology is flexibility. Multiagent technology enhances the adaptability of IT systems in two ways: it facilitates “external” maintenance and it increases their own capabilities to perform necessary adaptations by themselves. These are important contributions to the competitiveness of enterprises as they increase their capabilities to react appropriately to changes in their markets, to meet their customers’ demands better than before, and to take greater advantage of new market opportunities than would otherwise be possible. This contribution develops a formal framework for the description, and analysis of the flexibility of multiagent systems. It elaborates this framework in much detail and evaluates it on the basis of examples from the field of multiagent engineering.

Stefan Kirn

Application Examples I: Agent.Enterprise

Frontmatter

1. Agent.Enterprise in a Nutshell

Abstract

The manufacturing logistics domain is continually evolving towards ever more complex supply chain structures which call for increasingly flexible production capabilities. A concept is proposed which leverages agent technology’s capabilities to provide flexibility in such a complex environment while at the same time it does not interfere with the manifold interdependencies and individual behaviors of actors in modern enterprise networks. The Agent.Enterprise concept is demonstrated in a prototype implementation which integrates various multiagent systems (MAS) into a multi-multiagent system (MMAS). It provides integrated yet distributed and flexible supply chain management, from inter-organizational coordination down to detailed shop-floor level production planning.

Peer-Oliver Woelk, Holger Rudzio, Roland Zimmermann, Jens Nimis

2. Integrated Process Planning and Production Control

Abstract

This chapter deals with the application of intelligent software agents to improve information logistics in the area of process planning and production control. Therefore, enterprises will be able to fulfill the requirement of flexible, reliable and fault-tolerant manufacturing. Fulfillment of these requirements is a prerequisite for successful participation in modern business alliances like supply chains, temporal logistics networks and virtual enterprises. Thus, agent-based improvements of information logistics enable enterprises to face the challenges of competition successfully. Conducted research activities focused on the development of agent-based systems for integrated process planning and production control. They led to the “IntaPS” approach which is presented in this chapter.

Leif-Erik Lorenzen, Peer-Oliver Woelk, Berend Denkena, Thorsten Scholz, Ingo J. Timm, Otthein Herzog

3. Benchmarking of Multiagent Systems in a Production Planning and Control Environment

Abstract

Multiagent systems (MAS) offer new perspectives compared to conventional, centrally organized architectures in the field of production planning and control. They are expected to be more flexible while dealing with a turbulent production environment with its environment-immanent disturbances. In this chapter, a MAS is developed and compared to an Operations Research Job-Shop algorithm using a simulation-based benchmarking scenario. Environmental constraints for a successful application of MAS are identified and classified to be applied to next generation priority-rules based decision algorithms in MAS-based production planning and control.

Jan Wörner, Heinz Wörn

4. Distributed Hierarchical Production Control for Wafer Fabs Using an Agent-Based System Prototype

Abstract

FABMAS is a hierarchically organized multiagent system for production control of semiconductor wafer fabrication facilities (wafer fabs). The production control of wafer fabs is challenging from a complexity and coordination point of view. Semiconductor manufacturing involves one of the most complex manufacturing processes ever used. In this paper, we describe the application domain and major design decisions that lead to the FABMAS system prototype. A detailed discussion of the suggested software architecture of the agent-based system is included. Furthermore, we present the results of computational experiments that show that FABMAS outperforms dispatching based production control schemes that are currently in use. The paper also discusses some limitations and drawbacks of the suggested approach and identifies areas of future research.

Lars Mönch, Marcel Stehli, Jens Zimmermann

5. Supply Chain Event Management With Software Agents

Abstract

Operational fulfillment of supply chain processes in enterprise networks is regularly affected negatively by disruptive events. Event management promises to identify such problems in a timely fashion and significantly increase reaction time. A concept based on software agent technology is presented which enhances time and defect flexibility of supply chain processes. Evaluation of the concept indicates cost and cycle time reductions in multi-level supply chains which are not achieved by conventional approaches.

Roland Zimmermann, Stefan Winkler, Freimut Bodendorf

6. Trust-Based Distributed Supply-Web Negotiations

Abstract

This chapter presents a decentralized negotiation protocol for cooperative economic scheduling in a supply chain environment. These protocols are evaluated using software agents that maximize their profits by optimizing their local schedule and offer side payments to compensate other agents for lost profit or extra expense if cumulative profit is achievable. To further increase their income the agents have to apply a randomized local search heuristic to prevent the negotiation from stopping in locally optimal contracts. We show that the welfare could be increased by using a search strategy similar to Simulated Annealing. Unfortunately, a naive application of this strategy makes the agents vulnerable to exploitation by untruthful partners. We develop and test a straightforward mechanism based on trust accounts to protect the agents against systematic exploitation. This “Trusted” Simulated Annealing mechanism assures truthful revelation of the individual opportunity cost situation as the basis for the calculation of side payments.

Tim Stockheim, Oliver Wendt, Wolfgang König

Application Examples II: Agent.Hospital

Frontmatter

1. Agent.Hospital — Health Care Applications of Intelligent Agents

Abstract

In SPP 1083 the Hospital Logistics group studies the applicability of agent-based information systems in health care business scenarios by identifying problems, analyzing requirements, elaborating the state of the art of conventional and agent-based systems, specifying and designing multiagent applications, and evaluating their application. This chapter includes a survey of both the projects forming the group and their collaboration in order to integrate the systems designed by them into the agent testbed named Agent.Hospital. Therefore, two exemplary (hospital) processes are presented involving each project’s multiagent application. Also, the ontology OntHoS and agent infrastructure services used in Agent.Hospital are shown.

Stefan Kirn, Christian Anhalt, Helmut Krcmar, Andreas Schweiger

2. Artificial Software Agents as Representatives of Their Human Principals in Operating-Room-Team-Forming

Abstract

The scheduling of centralized operating theatres in large hospitals can be regarded as an archetypal cooperative decision problem. Multiagent systems (MAS) form an appealing paradigm for solving such problems. In a MAS-setting, each involved individual can be represented by an intelligent software agent that carries the specific constraints and the main preference-structures of his human principal. The scheduling can then be done by inter-agent negotiations, resulting in a cooperative solution, which optimizes “social welfare” and medical and organizational resource allocation simultaneously. For measuring human preference structures a concept based on conjoint analysis is introduced, that deduces individual utility functions suitable for inter-agent negotiations from human preference statements. Aggregation of individual preferences to find a final compromise schedule is then done by a distributed negotiation mechanism, based on the Nash-Bargaining-Solution of game theory.

Marc Becker, Hans Czap

3. Agent-Based Information Logistics

Abstract

Situated and context-sensitive information logistics surface as decisive requirement for supporting critical care units, because information relevant for patient treatment stems from heterogeneous as well as distributed data sources. Immediate treatment starts with an incomplete and almost empty array of information which fills continuously over time by examinations conducted by organizationally as well as geographically distributed departments. Agent technology and multiagent systems appear as a promising enabling technology to improve information logistics in intensive care units. However, an overall development methodology is required that enables an engineering process from the capture of know-how about clinical processes towards a model-based generation of multiagent systems. This contribution reports on an agile development methodology used for the design, implementation and testing of applications for agent-based information logistics.

Thomas Rose, Martin Sedlmayr, Holger Knublauch, Wolfgang Friesdorf

4. Agent-Based Patient Scheduling in Hospitals

Abstract

Patient scheduling in hospitals is a very complex task. This complexity stems from the distributed structure of hospitals and the dynamics of the treatment process. Hospitals consist of various autonomous, administratively distinct units which are visited by the patients according to their individual disease. However, the pathways (the needed medical actions) and the medical priorities (the health condition of the patients) are likely to change due to new findings about the diseases of the patients during examination. Moreover, the durations of the treatments and examinations are stochastic. Additional problems for patient scheduling in hospitals arise from complications and emergencies. Thus, patient scheduling in hospitals requires a distributed and flexible approach. To this end, a flexible, agent-based approach to patient scheduling is developed in this chapter. After a description of the addressed patient scheduling problem, the proposed mechanism for patient-scheduling is presented and evaluated.

Torsten O. Paulussen, Anja Zöller, Franz Rothlauf, Armin Heinzl, Lars Braubach, Alexander Pokahr, Winfried Lamersdorf

5. Adaptivity and Scheduling

Abstract

The structures in health care are currently changing. Clinical management and physicians have the obligation to both ensure quality of care and to work more cost effectively. The optimization of the system respecting these contrary goals is a big challenge. New information technology and computer applications like adaptive agent based assistance agents may be one way to optimize the system. Additionally organizational changes regarding resources or processes may also enhance the system. In many cases the effects of optimization ideas are difficult to foresee. This chapter describes the possibilities of multiagent simulation for experimentation and optimization of adaptive scheduling in hospitals. It presents a specialized agent based construction kit for hospital simulation and describes the results of realized example scenarios.

Rainer Herrler, Frank Puppe

6. Active, Medical Documents in Health Care

Abstract

Distributed and heterogeneous information systems can be observed in health care. In order to implement the vision of seamless health care, the boundaries of institutions need to be closed. Furthermore, information needs to be provided to the members of the health care team according to the principle of information logistics for the effective and efficient support of treatment processes. Since health care can be understood as a complex, adaptive system, an agent-based approach for an information system being capable of reacting flexibly to changes in its environment is an adequate solution. After the identification of characteristics of the health care domain the solution concept of active, medical documents is described, complemented by an analysis and development approach for a corresponding agent-based system.

Andreas Schweiger, Helmut Krcmar

7. Self-Organized Scheduling in Hospitals by Connecting Agents and Mobile Devices

Abstract

This chapter describes the conceptualization and realization of a real-time managed mobile information system in healthcare. The particular application addressed is a dynamic, self-organized scheduling of the treatment of patients. Building blocks for this project are locatable, interactive Personal Digital Assistants (PDAs) to connect medical staff and patients; physical resources are connected by locatable Radio Frequency Identification (RFID) chips. These physical objects are represented in the information system by software agents. The multi-agent platform EMIKA implements a negotiation-based schedule system to enable a dynamic planning process. The EMIKA-System has been developed to prototype level and functionally tested in a real-time laboratory [SaEM2002]. Lessons learned from the realization pertain to technical functionality and to privacy and security issues.

Torsten Eymann, Günter Müller, Moritz Strasser

Agent Engineering

Frontmatter

1. The Engineering Process

Abstract

Engineering highly flexible software systems for real-world applications on the basis of intelligent agents and multiagent systems is a challenging task. Conventional software engineering provides established methodologies and tool support. Additionally, knowledge engineering captures the necessary aspects of integrating knowledge in intelligent agents. However there is still a gap between software and knowledge engineering methodologies. State-of-the-art approaches of agent-oriented software engineering partially integrate these approaches. Nevertheless, challenges for the engineering process of agent technology remain open and therefore are addressed in this section on agent engineering.

Ingo J. Timm, Thorsten Scholz, Holger Knublauch

2. Requirements Engineering

Abstract

This chapter investigates requirements engineering (RE) for agent-based systems and in this context discusses the assumption that requirements engineering methods and tools are independent of the used technology. For this purpose, specific needs of agent-based systems that must be considered during requirements engineering are examined. As a possible approach for the fulfillment of the discussed needs, the capability of methods of goal-oriented requirements engineering (GORE) to support requirements engineering for agent-based systems is investigated.

Thomas Bieser, Hendrik Fürstenau, Stephan Otto, Daniel Weiß

3. Interaction Design

Abstract

Interaction is one of the core challenges in multiagent systems. It enables agents to share their knowledge, to do competitive or cooperative planning, coordination or bargaining, to interact with their principals, and to simple act. Interaction has to be restrictive enough to enable reliable system behavior and should be permissive enough to allow for flexibility or emergent behavior and performance. Obviously, the design of interaction differs from interface design in conventional software engineering significantly. Agents may also be designed for use in changing, respective unknown environments.

Karl-Heinz Krempels, Otto Spaniol, Thorsten Scholz, Ingo J. Timm, Otthein Herzog

4. Architectural Design

Abstract

This chapter introduces a reference architecture that provides a methodical framework for the implementation of software agents. The central concept is a layered architecture where each layer offers a well-defined service to the higher layers, and where each of the agent properties of I.1 is unambiguously associated with a single layer. The design method proceeds in three phases: A first phase examines the functional service characteristics in order to determine which responsibilities should be assigned to single agents and which to collections of agents. The second phase structures the individual agent into the layers along the non-functional properties. The third phase augments the layered structure by the interactive capabilities of agents.

Peter C. Lockemann, Jens Nimis, Lars Braubach, Alexander Pokahr, Winfried Lamersdorf

5. Semantics for Agents

Abstract

In the previous chapters design issues have been discussed with respect to architecture and interaction. These chapters are focused on techniques and methodologies. However, an explicit semantics of agent architectural and interaction behavior is required for the analysis and implementation of complex real-world systems. Additionally, the adaptation of a multiagent system to real-world domains requires the formal specification of knowledge. In this chapter, we introduce semantics from a general perspective with special focus on agent technology, e.g., semantics for communication. Finally, we introduce formal semantics for representation of intelligent agents and multiagent systems.

Thorsten Scholz, Ingo J. Timm, Otthein Herzog, Günter Görz, Bernhard Schiemann

6. Towards Dependable Agent Systems

Abstract

If an environment depends on the services of a multiagent system it should do so only if it can justifiably place reliance on this service. If so, the system appears to the environment reliable, or dependable. It is well-known that dependability should be designed right into a system rather than added as an afterthought. Particularly due to the high degree of distribution and the autonomy of agents, multiagent systems pose numerous and often novel challenges but also offer new opportunities to deal with dependability. This chapter examines the important issues and discusses how appropriate solutions can be associated with specific layers of the reference architecture of IV.4. Specifically, a distinction is made between unintentional and intentional failures, the former resulting in a suite of solutions referred to as error processing, the latter in measures called trust management.

Jens Nimis, Peter C. Lockemann, Karl-Heinz Krempels, Erik Buchmann, Klemens Böhm

7. Tools and Standards

Abstract

In this chapter tools, especially agent platforms, and relevant standards for realizing agent-oriented applications are presented. As there are a plenty of different agent platforms available the objective here is not to present an exhausting platform comparison, but to introduce meaningful platform categories, relate them to existing standards and illustrate them with typical representatives. The categorization helps to understand the existing heterogeneous agent technology landscape and is one integral part of a proposed selection method. This method reflects the fact that different problem domains may demand very different solutions in terms of the used methodology and underlying agent platform. It sketches the important steps that can be used to find a suitable methodology and agent platform fitting to the problem domain at hand.

Lars Braubach, Alexander Pokahr, Winfried Lamersdorf

8. From Testing to Theorem Proving

Abstract

Verification and validation of software systems are essential aspects in the software development life-cycle. However, verifying AI software is difficult as it suffers from non-determinism. In multiagent systems, this problem is increased by the known problems of verifying concurrent, distributed or object-oriented systems. On the basis of challenges for verification of multiagent systems, approaches for testing, runtime monitoring, static analysis, model checking, and theorem proving are discussed.

Ingo J. Timm, Thorsten Scholz, Hendrik Fürstenau

Evaluation

Frontmatter

1. Benchmarking of Multiagent Systems

Abstract

This chapter introduces benchmarking as a special form of evaluation and addresses problems and demands concerning the benchmarking of multiagent systems. It gives an overview of the evaluation concepts used in the German research program SPP 1083 for intelligent agents and realistic commercial application scenarios as well as examples for evaluation and benchmark studies for multiagent systems. The article provides basics for setting-up evaluation studies, regarding special concerns for the evaluation of multiagent systems. Moreover, the exemplary overview may serve as orientation for further evaluation and benchmarking of multiagent systems in realistic and commercial application scenarios.

Anja Zöller, Franz Rothlauf, Torsten O. Paulussen, Armin Heinzl

2. Simulation

Abstract

Multiagent simulation applies the concepts of multiagent systems to simulation. It is a perfect means to represent and examine emergent effects in distributed systems. Multiagent simulation models may be used to gain insights into system interdependencies, to make predictions and also for testing software systems. This chapter introduces the basic concepts and tools for agent based simulation and shows the possibilities of agent based simulation by means of the particular tool SeSAm. Extensions are presented that allow using multi agent simulation as an evaluation testbed for agent based application systems.

Rainer Herrler, Franziska Klügl

3. Legal Consequences of Agent Deployment

Abstract

The following chapter discusses various legal aspects concerning the implementation and usage of (software) agents. The underlying question if and how agents can issue legally effective declarations of intention is paramount in order to enable any reasonable usage of agents and, therefore, evaluated in greater detail in Section 3.2. Furthermore, the issue of how to deal with potential errors in agents’ declarations (Section 3.3), the requirements of agents’ signatures (Section 3.4) as well as liability (Section 3.5) and consumer protection issues (Section 3.6) are discussed. Finally, data protection issues and their implications on the usage of agents (Section 3.7) are scrutinized.

Tanja Nitschke

Backmatter

Titel: Multiagent Engineering
herausgegeben von: Professor Dr. Stefan Kirn
Professor Dr. Otthein Herzog
Professor Dr. Peter Lockemann
Professor Dr. Otto Spaniol
Verlag: Springer Berlin Heidelberg
Electronic ISBN: 978-3-540-32062-3
Print ISBN: 978-3-540-31406-6
DOI: https://doi.org/10.1007/3-540-32062-8