The Community for Technology Leaders

Guest Editor's Introduction: Internet-Scale Network Intelligence


Pages: pp. 38-40

There is no news in saying that Internet technology is revolutionizing all industries. From e-GM to thousands of nascent ".com" startups, a new way of doing business is well under way. The Web in particular has changed the face of commerce to something that clearly links buyers and sellers, establishing an interactive environment for delivery of goods and services. For manufacturing to exploit this interactive environment, however, certain key technical problems must be addressed. The theme articles in this issue of IEEE Internet Computing look at some of them.

Although the articles focus on manufacturing, they have broader applicability and close analogs to problems in electronic commerce in general. I will let the articles speak for themselves and focus here instead on describing what I believe are the key issues in manufacturing, especially as they apply to challenges and solutions relating to the Internet in general.

The very success of the Web has generated problems. For example, it has grown beyond the capabilities of Internet search engines to index it. More important to manufacturing and supply chains, Internet technology does not yet support a fully open market. In both search and interaction, consumers and producers must understand common terms and agree on some desired combination of single-site services. Some sites now provide simple agent technologies for notification and personalization, but there is much yet to do to make Internet connectivity more communicative, adaptive, and goal-oriented—in a word, more intelligent.


Internet technology is strongly focused on end-users and collaboration—perhaps most evidently in the Web. This focus is at odds with the manufacturing infrastructure of the first industrial revolution, which supported mass production and which was, and still is, dominated by centralized control systems. Just as the Internet is changing the face of commerce to a Web browser, it is driving a change in the infrastructure to support mass customization and immediate delivery of goods and services. This new industrial revolution will require the utmost flexibility in manufacturing and supply-chain operations.

The supply chain of the U.S.-based automobile manufacturers is now the world's largest extranet, Automotive Network Exchange. ANX represents a shift in supply-chain interactions away from "heavyweight" electronic data interchange (EDI) systems toward the ubiquitous and simpler Internet. ANX's first concern was to harden the Internet backbone, particularly for secure information exchange. ANX is now an established virtual private network, available though the many vendor firewalls supporting IPsec and Internet Key Exchange (IKE).

However, ANX still lacks the application and value layer for partners to coordinate dynamically with each other inside such secure tunnels.


IKE can be seen as a trend toward agent-based systems, which might also be used higher in the communications stack. Because the supply-chain partners are autonomous, IKE requires a primitive, form-based negotiation to connect them with each other. Even to establish a secure tunnel between two autonomous organizations, the IKE protocol specifies the use of proposals, which the automated policies of each party must understand and agree with.

IETF's work on policy definitions and Quality of Service (QoS) negotiations is another area where systems require more intelligence as they become more autonomous and their options and conflicts become more complex. Particularly for the historically conservative manufacturing industry, local control and systemwide guarantees of behavior are absolute requirements for acceptance. Policy definitions, being simpler than full-blown agents, will provide a good entry point, but agents will be necessary for greater flexibility.

Of course, ontology-like catalog systems and online ordering services such as FastParts are already becoming well established on the Internet.

Simple agents—such as notification agents—are common. Automatic pager notification of flight changes can be requested on some travel sites. Although such function is simple and not particular to Internet technology except in its Web access, notification and distributed coordination can be valuable when applied to engineering and logistic changes.

More substantive intelligent agents, such as those specified by the Foundation for Intelligent Physical Agents (FIPA), are still emerging, but almost all platform implementations are based on conventional Internet technologies: Java, XML, IIOP and/or TCP, and often LDAP as well.

A complete list of FIPA-related software and field trial results is available from its Web site (see the sidebar, " FIPA and OMG Work in Product Design and Manufacturing").


Given its conservatism, manufacturing is paradoxically also one of the oldest and strongest areas of agent research. Contract Net has been applied to dynamic scheduling for more than a decade. The European Community and the U.S. National Institute for Standards and Technology (NIST) have funded many pilot projects and consortia to advance both the research and commercialization of agent-based manufacturing.

Some manufacturers have adopted principles of agent design, and commercial agent providers such as Gensym and Object Space are committed primarily to their manufacturing-support business. Again, Java and other Internet-relevant technologies are common ingredients in most single-point applications of agents to manufacturing. A multi-vendor, international system has not yet appeared, but the global arena of open standards is where the true potential of agent-based manufacturing lies.


Developing distributed software systems for new types of organizations on the Internet is capturing the imagination of developers and users of agent-based systems. This issue of IEEE Internet Computing addresses this subject in the manufacturing industry, which provides an excellent testbed for issues of autonomy, distributed knowledge, globalization, and massive scale.

Whether or not manufacturing per se is your particular interest, we hope you enjoy how the implementations reported here address this future of the Internet. Manufacturing will continue to be a most challenging case for e-business, requiring Internet intelligence running at Internet scale.



Fernando Zubillaga-Elorza and Charles R. Allen

A Java-based set of electronic design automation tools can be used to simulate conceptual designs. Users can interact with the design through a Web browser, and designers can use the feedback to refine the design specification early.


Padmanabh Dabke

A collaborative infrastructure supports large-scale enterprise integration by representing product data and organizational resources as components and by maintaining high-level process specifications as agents.


Carmen M. Pancerella and Nina M. Berry

Sandia National Laboratories is building a multiagent system on top of its existing Product Realization Environment (PRE), a framework for enterprise integration.


Nenad Ivezic, Thomas E. Potok, and Line Pouchard

The transition from traditional, centrally scheduled "push" systems of production scheduling to customer-driven "pull" systems and globally synchronized "takt" scheduling is a complex, enterprise-wide process. MABES can help.


Albert D. Baker, H. Van Dyke Parunak, and Kutluhan Erol

The Internet supports global communication between customers and manufacturers, but the physical realities of manufacturing require something more.

FIPA and OMG Work in Product Design and Manufacturing

The Foundation for Intelligent Physical Agents develops specifications for agent directories, communications, wrappings, ontologies, human interaction, and learning. These specifications provide an important basis for coordinating stakeholder efforts to integrate virtual enterprises across distributed, multicompany, semi-autonomous teams.

The FIPA Web site is at

The Object Management Group and its Agent Working Group are in liaison with FIPA, working on the integration and specific reification of FIPA agents with CORBA objects. Additionally, OMG's Manufacturing Domain Task Force is defining the activity-resource workflows and domain object ontologies, with which such agents would work.

The OMG Web site is at

About the Authors

Manuel Aparicio IV is co-founder and chief technical officer of Saffron Technology. Previously, he was chief scientist of IBM's Center for Intelligent Agents and Knowledge Management, where he codeveloped IBM's first commercial agent system in 1993, and led the development of many commercial contracts, particularly in manufacturing and telecommunications. Aparicio has patented several learning algorithms and agent methods, including human-agent interaction. He is the North American Director to the board of FIPA. He received a PhD in psychology from the University of South Florida as a National Science Foundation Minority Fellow in 1988.
60 ms
(Ver 3.x)