FIPA - Intelligent agents from theory to practice

Manuel Aparicio - IBM, USA; Leonardo Chiariglione - CSELT, Italy;
Ebrahim Mamdani - Imperial College, UK; Francis McCabe - Fujitsu, USA;
Richard Nicol - BTLabs, UK; Donald Steiner - Siemens, Germany; Hiroki Suguri - Comtec, Japan



Intelligent agents belong to a computer science area that is rapidly achieving maturity. They promise to have a huge impact on a wide range of application domains, particularly telecommunications. The Foundation for Intelligent Physical Agents (FIPA), an industry consortium established in 1996, has already produced specifications for various aspects of agents having in mind all industries likely to be affected by the technology. Several implementations have been independently developed and tested for interoperability. The paper also describes a selection of application domains where FIPA agents can be successfully introduced.


1. Agents, the new software paradigm

Today, unless special care has been taken in the design of the code, two software programs cannot interoperate. The promise of agent technology is to move the burden of interoperability from software programmers to programs themselves. This can happen if two conditions are met. First a common language, called Agent Communication Language (ACL), must exist to enable software to recognise the intention behind a request. Second an architecture must also exist where a piece of software can describe its capabilities and needs and can discover the same for other pieces of software.

Further to that, the ACL must be expressive enough to enable complex forms of interaction, not simply at a syntactical level, as in the case of CORBA but, mainly, at the semantic level where a community of agents, i.e. pieces of software endowed with such advanced features and residing in an appropriate environment, may request information and proposals, delegate a task, compete for it, enter into an auction, negotiate the terms of a contract, etc. These are patterns of interaction common to humans but unusual for traditional software. But these are patterns that might open the way to new kinds of applications and to new kinds of software assistance, and in general to new effective ways of reasoning about and exploiting that distribution of data, control, intelligence, expertise and resources that is a characteristics of a large share of today's applications.

Intelligent agents thus, provide a powerful paradigm that has already found some practical applications. When agents interact with other agents, the result is a networked intelligence that co-operate in the delivery of useful and far reaching new services. This kind of "social agents" need standards. As long as the different agents all come from the same source there are no interoperability problems, and the features of these agents are completely within the freedom of the vendor. However, the scope of use of the technology is severely limited and unlikely to create that self-enhancing phenomenon that has been witnessed with the WWW and GSM telephony. The absence of standards shared by multiple vendors may help to explain the limited success that some agent-based applications have had so far.

FIPA was born in 1996, not as a reaction to the then current problems but as a plan to address future imperatives. Even after 3 years, we can only now say that the market demands are maturing and that FIPA technology is becoming gradually commercially available, even at the time of writing this article. Some may say that the demands and solutions for intelligent agent technology still need another year or so to mature. On the other hand, the market forces in telecommunications, particularly in Europe, were so pressing even as early as 1996 that several companies with different but aligned interests founded FIPA to prepare for this impending future. In some estimates, 3-4 years is a "long haul" to plan and build an information technology, but the opportunities for this future are enormous. The long awaited information age has finally begun, primarily around the Internet, and is maturing rapidly. We are in the first stirrings of a new decades-long industrial revolution. The forces of this revolution will require standards, especially standards for intelligent agents.

European Telecommunications was a primary founder of FIPA because of increasing international and multi-vendor pressures. By definition, any long-distance provisioning is multi-national and multi-vendor, and globalisation is increasing the need for such long distance provisioning. Geography is one factor that determines whether markets will be consolidated by a dominant few (classic market leaders) or differentiated to local and more niche providers. As a mature industry understanding such forces, the providers understood that provisioning was becoming more and more into a commodity, the value of which should be negotiated in real-time depending on current supply and demand. The properties of intelligent agents, such as being autonomous, reactive, communicative, and adaptive, seemed to match this growing trend in how different, autonomous companies could more closely compete/cooperate with each other in a beneficial way.

Manufacturing provides another industry example. This industry was and still is largely monolithic. The industrial revolution was characterised by mass production, such as Henry Ford's single colour (black) Model T. More recent pressures have emphasised product quality and then execution speed. The next pressure, much like in telecommunications, will be for flexibility in the manufacturing process. While the established large manufacturers also include other primary industries such as steel and rail, to provide the support infrastructure a completely reverse trend is now taking place. This involves a move towards "manufacturing-less" manufacturing. This recognises that a manufacturing companiy’s most valuable asset is its design and brand, while the actual building of components and elemental parts can be contracted out. Actual manufacturing is becoming a commodity, with supply-chains of consumer-producer becoming much more dynamic and even overlapping in co-operation/competition. Large companies still dominate the supply-chain, typically enforcing strategic partnerships such as for strategic purchasing, but online auctioning for parts and small agile job-shops are beginning to emerge and likely to become more important in the future. Thus the need for speed and flexibility between autonomous manufacturing companies is also a ripe opportunity for intelligent agent being used in negotiation and co-ordination.


2. Technology availability and market necessity

Real businesses also extend beyond any one industry. The telecommunications interest in agents goes well beyond bandwidth provisioning. All industries will be affected by intelligent agents they come to rely more and more on telecommunications. For instance, the world’s largest extranets, for the US’s automaker's supply chain, is managed by the area telecom provider. In Europe, the co-ordination of international travel (business or pleasure) is seen as another ripe area for the exploitation of intelligent agents. Aside from the same multination and multi-vendor situation as already discussed, travelling in Europe is more difficult because of the variety of travel means and choices. Such plans must often include planes, trains, automobiles, and even walking. All these means are often under changing conditions of service. Nor is the mode of transportation the only part of the task of planning a journey. Hotels, business arrangements, personal pleasure activities, financing, and several other closely aligned industries in "travel" also need to work dynamically with each other, not just for optimal pre-trip planning, but for on-trip re-planning. Standardised agents can facilitate such a multi-vendor, multinational, and even multi-industry "bouquet of services".

The same is true for Manufacturing. In this case, the "bouquet" of required service integration is even large, including travel within it. From consumer marketing to product design to manufacturing per se, to the logistics of delivery all along the way, co-ordination and notice-of-change must percolate through an enormously complex set of relationships and dependencies. The supply-chain of airfreight for instance is already becoming an overlap of people and cargo. Air-carriers would be more efficient to dynamically provision their cargo operations, but this will require real-time dynamic co-ordination between manufacturers, carriers, and the telecommunication provisioning (land and wireless) between all these organisations -- and most importantly, people.

Global communications and travel, mass customisation and personalisation, and Internet access to information and commerce is spoiling the consumer. Consumers are not just demanding but more and more simply assuming that they can find and get the options they want. This demand will favour the companies that supply it, which makes the assumption even stronger and expectations cannot go back. Rotary phones to call anyone else were once a wonder, but given the speed of tone dialling, rotary dialling is perceived as excruciatingly slow.

Anywhere, anytime consumer access to the Universal bouquet of information and services is the new goal of the information revolution. The Internet is becoming the expectation and solution toward this goal, made possible because its is standard, almost by definition. As mentioned above, standardisation makes choice possible. Component standardisation in any one domain makes plug-and-play possible, but the scope of Internet standards makes the scope of choices extreme.

Audio/video entertainment is also facing the same problem. Consumers with different tastes are demanding an ever increasing variety of niche offerings. The technology for delivery of any such media is becoming well established in satellite, digital cable, and other channels, but the scope of choices is likewise becoming extreme.The successes of the Web and traditional broadcasting are creating new problems in this regard. Too much information is no information if it cannot be found. The growing Web is also growing a disincentive to use the Web. Search engines are less and less able to index it and their perceived accuracy is diminishing as more and more common users log on and expect more. Even very sophisticated users are becoming more frustrated by search results that are dated, irrelevant, and simply too voluminous.

We cannot go back. Everyone does not want a black Model T. Portals are emerging as one solution to this problem. Given a particular subject like music or health, a portal will advertise its site in order to attract notice and usage -- not search engines. But this is temporary, at least as the single solution. Internet and intelligent agent technologies will allow for the more natural bazaar of many differentiated niche services in the universal bouquet. Intelligent agents will find, broker, negotiate, and co-ordinate consumers and suppliers, making the Web more intelligent..


3. The FIPA standards for agents

The Foundation for Intelligent Physical Agents (FIPA), established in 1996 in Geneva, is as an international non-profit association of companies and organisations which agree to share efforts to produce specifications of generic agent technologies.

At a ground-breaking meeting at London in April 1996 and a follow-up meeting at Yorktown Heights, NY in June 1996, FIPA adopted the above principles and started work along the following process:

  1. Agreement on and adoption of basic terminology and reference architecture
  2. Identification of potential areas of agent technology suitable for standardisation (long list)
  3. Identification and selection of application areas to be targeted
  4. Selection of specific areas of agent technology to be targeted (short list)
  5. Production of normative specifications for the technology and informative specifications for the application of the technological specifications to the application areas. Based upon responses to publicised call for proposals, specifications are drafted and made public for comments. After several iterations of comments, the FIPA members formally ratify the specifications.
  6. Testing of the specifications based on the applications (field trials). This is primarily the responsibility of individual organisations, and is usually carried on outside of FIPA itself, though, as we will see below, specific interoperability experiments are supported by FIPA.

Over the ensuing meetings (held at quarterly intervals for one week at alternating international locations) FIPA continued this work, leading to the first series of specifications – FIPA 97 – ratified at Munich in October 1997. The normative (technological) parts of FIPA 97 comprise the following:

Spec 1 - Agent Management – services (e.g. yellow and white pages) providing means for agents to find each other based on their capabilities and to exchange messages with each other.

Spec 2 - Agent Communication Language – high-level, speech act-based language (FIPA ACL or Agent Communication Language) used in messages among agents to support information exchange, task distribution and negotiation as well as predefined interaction protocols composed of the basic message types (e.g. contract net or common auction protocols).

Spec 3 - Agent/Software Integration – services supporting integration of existing legacy software systems into an agent system.

The informative (application) specifications of FIPA 97 comprise:

Spec 4 - Personal Travel Assistance – integration of heterogeneous travel services to provide automated, individualised support for planning and executing complete (door-to-door) travel itineraries.

Spec 5 - Personal Assistant – individualised office workflow support, in this case targeting scheduling group meetings taking meeting places and travel time into consideration.

Spec 6 - Audio/Visual Entertainment & Broadcasting – support of automated individualised programming of broadcast services.

Spec 7 - Network Management & Provisioning – automated configuration and negotiation of virtual private network services, taking the interests of the human users, service providers and network providers into account.

In the following year FIPA improved on Specs 1 and 2 and targeted new areas, resulting in the FIPA 98 specifications, ratified at Durham in October 1998:

Spec 8 - Human-Agent Interaction – services for managing multi-modal (e.g. text, speech and graphic-based) dialogues with humans and user personalisation services including maintenance and adaptive construction of user models.

Spec 10 - Agent Security Management – identification of the key security threats in agent management and specification of facilities for securing agent-agent communication via the FIPA agent platform including agent-agent interaction, agent configuration, agent-agent platform interaction, user-agent interaction and agent mobility.

Spec 11 - Agent Management Support for Mobility – services supporting software agent mobility using the FIPA agent platform, taking existing standards in this area (e.g. OMG’s MASIF) into account.

Spec 12 - Ontology Service – services enabling agents to manage explicit, declaratively represented ontologies, including discovery of public ontologies, identification of relationships between ontologies and of shared ontologies, as well as translation between ontologies.

Spec 13 - FIPA 97 Developer's Guide –a companion document to the FIPA 97 specifications, clarifying areas of specific interest and potential confusion.

FIPA is currently working on the FIPA 99 specifications, expected to be ratified in October, 1999. These will include (as of July 1999):

Spec 0 - FIPA Architecture – formalised (UML-based) description of the components of the overall FIPA architecture and their relationships and architectural guidelines for constructing agents and agent platforms.

Spec 14 - Nomadic Application Support – mechanisms supporting the wide class of applications which call for mobility of the user and associated computer systems. This includes message transport and appropriate encoding of message content (including ACL messages) over low-bandwidth and often disconnected communication channels.

Spec 16 - Agent Message Transport – mechanisms (including message transport service) for dealing with the wide variety of communication protocols which may be used between agents (e.g. TCP/IP, HTTP, IIOP, SMTP, etc.) as well as a library of specific communication protocols for use by FIPA agents.

Spec. 17 - Agent Naming – a flexible scheme of naming agents supporting globally unique identifiers, nicknames and mechanisms for resolving message transport addresses from the names.

Spec. 18 - Agent Content Languages – requirements on languages to be used in the content of ACL messages, as well as a library with specifications and encoding schemes of suitable content languages (e.g. FIPA SL, KIF, constraint languages and XML-based languages)

In addition, FIPA 99 is extending Spec 2, Agent Communication Language, with an enhanced library of message types and interaction protocols, as well as an extension of UML to formally describe interaction protocols.

The current ratified and draft FIPA specifications can be found on the FIPA web site,, and the mirror sites in Japan,, and the USA,


4. Implementations and interoperability trials

As of August 1999, fourteen organisations, both members and non-members [1] have announced implementations of FIPA-related software. The organisations are Broadcom, British Telecom, Comtec, CSELT, Fujitsu, Fujitsu Laboratories of America, GMD FOKUS, IRST, University of Electro-Communications, Nortel Networks, Saffron Technology, Siemens, The Nesstar Project and SPAWAR Systems Center. The degree of implementation differs among the systems: some of them are partial adoption of FIPA technology in the existing applications; some fully implement a part of the specifications from the scratch.

One of the most successful implementations of FIPA specifications can be seen in FACTS (FIPA Agent Communication Technologies and Services, European ACTS Project AC317) [2]. The FACTS project implements FIPA-compliant agent platforms (Specs 1, 2 and 3) and on top of them, three applications: personal travel assistance (Spec 4), audio-visual entertainment and broadcasting (Spec 6) and network management and provisioning (Spec 7). FIPA has also started by its own expense to implement a meeting scheduling application based on the Spec 5 scenarios [3]. The source code of the agent platform and the meeting scheduling application developed by FIPA will be made public as an open source package.

Open source movement is getting popular these days. Although the spirit of sharing the source code is as old as computer itself, commercial importance of the free software has not been properly recognised until recently. Among the fourteen organisations listed above, Comtec and Nortel Networks have already made the source code of their implementations open. Many other companies and institutes are seriously considering to do so..

The FIPA Statutes clearly say that "The purpose of FIPA shall be pursued by: identifying, selecting, augmenting and developing in a timely fashion specifications of generic agent technologies that are usable across a large number of IPAs [Intelligent Physical Agents] and provide a high level of interoperability with other agents-based applications." [4] But, why do we have to take care about the interoperability for all the specifications? It is important to remember that FIPA specification is very tolerant to subset and extension. In fact, one of the basic policies of FIPA technology is that, in most cases, implementors may choose subsets of the specification. Extension is also allowed as long as it does not conflict with the specification. For example, 'not-understood' is the only communicative act that every agent must understand. You can use any content language inside the communicative acts. Selection of the appropriate subset or superset is determined by the requirements of the application.

This principle is intended to make it easy to design and develop a flexible application systems. On the other hand, if each developer uses an arbitrary subset or superset of the specification, these agents would not be able to talk to each other. Therefore, guaranteeing the interoperability of the heterogeneous agents has a critical importance for the development and deployment of agent-based applications on the Internet.

FIPA conducted the first interoperability trials in the twelfth meeting in Seoul, Korea, on January 1999 [5]. In the first trials, four companies (Broadcom, Comtec, CSELT and Siemens), with different hardware, operating systems, Object Request Broakers (ORB) and languages, connected their agent platforms in a single LAN to test the interoperability and compatibility. Under test were basic agent management functionality and simple application scenarios of appointment scheduling. The tests were successful in most cases. It was amazing that the systems, which were separately developed by different companies on different platforms with different programming languages, connected so smoothly for the first time. Currently trials are being conducted on a public agent platform set up on the Internet [6] so that anybody can try to connect to the platform at any time.


5. FIPA-enabled applications

The range of possible applications for FIPA-compliant agent systems is limited only by the imagination. We envisage agent technology being useful in many if not most of the kinds of applications of distributed computer systems seen today. However, one particular hallmark of FIPA agent applications is likely to be `intent'. Intelligent agent technology is particularly appropriate in situations where it is important to be able to be explicit about the intentions of people in their computer-mediated interactions.

In that context, we can identify some classical applications where intelligent agent technology and FIPA compliant technology is likely to be critical

Enterprise applications

Smart documents.

A smart document combines the data aspect of internal and external documentation (such as a purchase order, or a patient's medical record) with an explicit intention (such as fulfilling the order, or completing a medical diagnosis). The explicit intention may be encapsulated in an intelligent agent that guides the document between automatic and human agents in the course of fulfilling the document's intention.

Goal-oriented enterprises.

A goal-oriented enterprise can be considered to be a slightly formalised version of today's project team -- incorporated as a separate legal entity. However, such a team may be very dynamic and created on a single-use basis -- once the team's objectives are met the company may be dissolved.

Members of such a company are likely to be involved in many simultaneously. Conversely, the `employees' of such a company may be actually employed elsewhere. The goal oriented enterprise is in reality formed from the collaboration of a number of other legal entities --which do not abrogate any of their prior rights and commitments.

Managing such goal oriented enterprises is only possible with the considerable help of the Internet, and with specialised agent-oriented tools to allow people to manage their own time and to allow managers to manage partners located in many different organisations.

An early prototype of such goal oriented enterprises can be seen today in the many collaborations sponsored by the EC within ESPRIT, ACTS and so on; and also by the US government in DARPA sponsored projects.

Role and personnel management.

In today's organisations, people change function and role very rapidly. Keeping up with such rapid change becomes impossible in any but the smallest organisation. Agent technology can be used to provide formal tools that manage people's roles in organisations and also controlling access to the resources available to people as their functions change in the organisation.

Such formal tools would also allow new capabilities such as role-based communication: I need the person delegated by the CEO to look after Asia to see this memo. Such an address is impossible today because conventional email or telephone requires that the sender knows the email or telephone number of the actual person receiving the message.

Inter-Enterprise applications

In addition to intra-enterprise applications for agent technology, we see a number of interesting possibilities for inter-enterprise applications. Here the role of open standards is critical.

Market making for goods and services

One very large area of inter-enterprise collaboration is simply the market. FIPA agent technology can be used to automate simple markets. Members participating in an automated market can leave some of the drudgery to agents. These agents are able to participate in an open market because there is common understanding of terms of communication and there is a common understanding of the commitments entered into. Both of these are critical pieces of FIPA technology.

Brokering of the above

In addition to simple market making, there is scope for the combination of services. Combined services may be anything from virtual catalogues that combine the catalogues of competing vendors to virtual services that combine different types of services to form new services. For example, a language translation service can be combined with a new provider service to enhance the capability of both.

Such services can often be combined using brokers and brokering services. A broker `brings together' a customer and a provider where neither may have been aware of the other.

Team management

The distinction between project teams located within an organisation and collaborations between organisations is blurring. Support for team management will have to be extended to include teams spread across organisations as well as within them. Such teams may be distributed globally as well as culturally.

To support inter-organisational teams requires strong standards that support the reliable communication of meaning as well as form of messages.

Process control

One of the most promising areas of applications of intelligent systems is in the control of widely distributed physical systems where there are many competing requirements.

Intelligent buildings

For example, the flow of heat and/or cooling in large buildings, represents a compromise involving the competing requirements of all of the residents of the building. Intelligent agent systems can be used to represent those users' requirements to the intelligent building.

Large scale plant and equipment

On the other hand, some structures cannot properly be controlled in a centralised way; either because they are very large or because their ownership is also distributed (such as the various networks that make up the global Internet). Intelligent agents that `speak' standardised languages (such as FIPA's Agent Communication Language) can be used to help coordinate these large structures.


Industrial robots have long been promised as a way of gaining unlimited productivity. Unfortunately, programming robots to be flexible and responsive has proved to be much harder than originally conceived. However, intelligent agent technology can assist in building robots -- both at the micro level and the macro level.

Many of the hardest tasks for robots involve activities that are trivially simple for humans: such as picking up delicate items without breaking them or throwing a ball to land in a well defined target. However one promising approach to the construction of flexible robots involves assigning agents to different parts the robot -- such as individual joints in a robot arm. The robot achieves its global task be requiring the collaboration of these smaller micro-agents.

Using standardised models of communication between these micro-agents allows teams and manufacturers to specialise in different parts of the construction of industrial robots.

At the macro-level some task cannot be accomplished by a single robot. For example, the managing of goods in a warehouse cannot easily be done with a single centralised computer system. Such an application requires the cooperation of many different types of system -- from planning the movement of goods, to the internal storing and retrieval of goods to the simultaneous loading and unloading of trucks arriving at the warehouse.

Multi-agent systems, and in particular FIPA-compliant multi-agent systems, provide an architectural platform for building collaborative and cooperating systems such as the warehouse. Having standards in this area allows different manufacturers to specialise in different kinds of robot systems (automated goods movement vs. long range and short range planning).

Personal agents

In addition to the industrial and commercial applications outlined above, there are many more personal possibilities for intelligent agent technology. Once again, the key attribute that is captured in an agent in intention. In the are of personal agents, this means that the automatic system is able to relate to the individual user in terms of his or her intentions.

Email and news filters

A classical application of agent ideas -- if not of actual technology -- is in the area of information filtering. email and news filters allow an individual to control the flow of information to help avoid information overload. In the case of a filter, the intention is clear: I want to be able to see all interesting news about X; or I don’t want to see any email from people promising yet another get rich quick scheme.

Personal management

Managing one's own time becomes more difficult as the demands get more intense. Unfortunately, there is every possibility that this will get worse not better. Personal management agents -- such as time management or travel management agents -- can help to automate some of the low-level tasks and help to ensure that personal constraints are satisfied where possible.

Personalised help

In addition to managing one's resources, managing the everyday complexity of life also gets more difficult. Here computers are at least as much the problem as the solution -- especially given the hard to use nature of most computer systems. Help agents that are able to take into account the actual pattern of use of computer systems and to tailor their advice based on that are yet another potentially fruitful application domain for intelligent agents.


The telecommunications industry is undergoing massive change. Growing deregulation, increased competition, and the globalisation of business are all influencing the shape of the modern communication's company. The "new-wave telco" needs to respond dynamically to the changing requirements placed on it by both customers and new markets.

We discuss a number of examples where agent technology will influence the telecommunications industry, its management systems and communication services.

Telecommunications management systems

The challenges facing telecommunications management (TM) are many. Costs have to be kept to a minimum to maintain competitiveness whilst the quality and effectiveness of the management systems has to improve to differentiate service. The flexibility of the IT infrastructure supporting TM is therefore pivotal to the success of a modern telecom organisation. Two areas in particular where this is the case is network management and service management.

With the opening of telecommunication markets bandwidth has become a tradable commodity.. Arbinet Communications Inc. [3], is offering an internet trading environment for network capacity. They are also facilitating the trading in futures and options in bandwidth. Although currently manual, software agents will eventually automate much of the negotiation for bandwidth either for end consumers or in the inter- carrier market. Already agent systems have been developed to provide such services, [4].

Agent-based process management systems [5] combine agent technology and distributed computing platforms. They take a service-oriented view of managing processes where different collections of activities are offered as services by contributing departments. Such systems reflect the inherent distribution of responsibility in large organisations, and make the management of an organisation transparent to the logical or physical structuring of its components. Similar systems have been developed to support inter-organisational (or supply chain) processes, [6]. Such agent systems represent an infrastructure for instant virtual enterprises which are dynamically configured to suite different needs, [7].

Allowing software agents to represent the interests of autonomous departments or business units allows organisations to adapt and evolve with minimal disruption. New services or tasks can be defined incrementally, without the need to re-design an entire distributed system.

Communication services

The portfolio of services on offer to customers of telecommunication companies is ever widening with increasing convergence of media, telecom and computing industries. As well as traditional telephone and fax services, internet (incl. knowledge management), LAN interconnect, mobile communications, e-business, e-commerce services are becoming mainstream products and services. Agents are well suited to supporting such services.


6. Conclusions

    In listing these numerous applications for intelligent agents it is important to remember that we are NOT talking about applying some kind of magical smart bullet to solve all known problems. Nor are we talking about reinventing all computer applications in an intelligent world.

    Instead, the main message is that the architectures involved in building `cognitive agents', and the infrastructure needed to support standardised communication between them allow numerous new possibilities that build on existing expertise and existing solutions.

    The main reason why we can see this potential is because FIPA's model of communication allows computer programmer to reason about the meaning of messages as well as their form. This additional capability makes it possible for the first time to allow computer applications to communicate in a much `stronger' way than simple text messages. FIPA's ACL is stronger in this regard than the W3C's new standard XML which allows the form of a document to be well defined but does not allow the semantic content of the document to be understood by computer applications.

    In each of these application areas the need for FIPA is apparent. The highly interactive nature of multi-agent systems, where agents negotiate for bandwidth, products and services, or where they co-operate to create instant virtual enterprises, point to the need for consensus on agent interfaces in order to support interoperability between different agent systems. The completion and adoption of the FIPA standard is a prerequisite to the commercialisation and successful exploitation of intelligent agent technology in this domain.

    FIPA has set itself a significant challenge in specifying a consortia standard where system inter-operation is at such a knowledge rich level. If FIPA achieves its goal and agent developers adopt the FIPA specifications, agent technology will represent a powerful software engineering paradigm for the next millennium.

8. References

  7. R.G.Davison, J.J. Harwicke, M.D. Cox, "Applying the agent paradigm to network management", BT Technology Journal, Volume 16, Number 3, July 1998.
  8. Finin T., "knowledge sharing KQML, KIF and Ontologies", tutorial, Practical Application of Intelligent Agents and Multi-Agent Technology, London, April 1997.
  9. ARBINET Communications introduces on-line trading of bandwidth derivatives, KCSA Website,, or May 1999.
  10. Miller M., Krieger D., Hardy N., "An Automated Action in ATM Network Bandwidth", in Market-based Control, ed. Scott H. Clearwater, World Scientific, 1996.
  11. O'Brien P.D., Wiegand M.E. , Agent based process management : applying intelligent agents to workflow, Knowledge Engineering Review, Vol. 13:2, September 1998.
  12. Shepherdson J.W., Thompson S., Odgers B.,"Cross Organisational Workflow Co-ordinated by Software Agents",WACC '99 (Work Activity Co-ordination and Collaboration) Workshop Paper, February 1999
  13. Robert J. Laubacher, Thomas W. Malone, and the MIT Scenario Working Group , "Two Scenarios for 21st Century Organizations: Shifting Networks of Small Firms, or, All-Encompassing "Virtual Countries"?",
  14. Soltysiak S.J., Crabtree I.B., "Automatic learning of user profiles - toards the personalisation of agent services", BT Technology Journal, Volume 16 No. 3, July 1998.
  15. N.J.Davies, R.S.Stewart, R.Weeks, "Knowledge Sharing over the world wide web", ", BT Technology Journal, Volume 16 No. 3, July 1998.
  16. L.Foner, "A Multi-agent referral system for matchmaking", Proceedings of First International Conference on the Practical Application of Agents and Multi-Agent Systems", London, 1996.
  18. ACSES
  19. Amazon
  20. Personal logic


9. Biographies

manny.jpg (65545 bytes)

Dr. Manuel Aparicio is co-founder and Chief Technical Officer of Saffron Technology. Previously as Chief Scientist of IBM's Center for Intelligent Agents and Knowledge Management, he consulted and led the development of many commercial contracts particularly in manufacturing and telecommunications. He is the North American Director to the board of FIPA and was a past director to the board of Agent Society. He received a PhD in Psychology from the University of South Florida as a National Science Foundation Minority Fellow in 1988 and now has over 15 years of experience in machine learning and intelligent agent technology.

He co-developed IBM's first commercial agent system in 1993 and has patented several learning algorithms and agent methods, including human-agent interaction. He has written articles, edited books, and presented talks covering both the technology and application of learning and agents.

Readers may contact him via email at

leonardo_in_office8.jpg (15548 bytes)

Dr. Leonardo Chiariglione is Head of the Multimedia Services and Technologies Research Division of CSELT, the Corporate Research Centre of the Telecom Italia Group. Outside his company he acts as "Manager of Technology Convergence" in his capacity as Convenor of the Moving Pictures Experts Group (MPEG), President and Chairman of Board of the Foundation for Intelligent Physical Agents (FIPA), originator of the Open Platform Initiative for Multimedia Access (OPIMA) and as Executive Director of the Secure Digital Music Initiative (SDMI).

Dr. Chiariglione obtained his Ph.D. from the University of Tokyo, and graduated in Electronic Engineering from the Polytechnic of Turin.

Amamdani.jpg (8220 bytes)

Professor Mamdani has had a 30 year long academic career at London University. His research interests have been in the area of machine intelligence where he his known for his contributions on uncertainty management systems, fuzzy logic and software agent based systems. In particular his work on Fuzzy control has been exploited in a range of consumer products from cameras to automobiles.

In 1995 he joined Imperial College, London to take up a newly created Chair of Telecommunications Strategy and Services endowed by Nortel Networks and the Royal Academy of Engineering. He is a Fellow of the Royal Academy of Engineering as well as a Fellow of the Institute of Electrical and Electronic Engineering.

frank.jpg (9365 bytes)

Dr. F.G. McCabe is a Senior Researcher at Fujitsu Labs of America. He is researching into the computational frameworks that underlie software agents and is developing programming languages and tools for building such systems. This includes communications infrastructure as well as high-level concurrent symbolic programming languages and agent application frameworks.

He has been involved with FIPA since 1998, and is also active in the Agents working group within OMG.

He has also worked extensively in the Logic Programming field, developing several variants of the Prolog programming language.

nicol.jpg (26720 bytes) Dr Richard Nicol has been with BT labs since 1970 , he spent many years involved with video coding research, technology, products and multimedia systems. He led several of the international formative standards groups on video compression.

For 3 years he ran BT's research activities with the Universities.

He is now BTs Head of Research.

In addition to being a founding director of FIPA, Dr Nicol is a Chartered Engineer, a Fellow of the Institution of Electrical Engineers, a member of the IEE Electronics and Communications Divisional Board, and a visiting Professor at the Universities of Essex and Lancaster.

don.gif (10899 bytes)

Dr. Donald Steiner has been involved in Distributed Artificial Intelligence and Multi-Agent Systems research for the last 10 years. He was Siemens' technical manager as well as a workpackage manager of the Siemens-led Esprit II Project IMAGINE, Integrated Multi-AGent INteractive Environment. The results of this project led to the design and implementation of the Multi-agent Environment for Constructing Cooperative Applications (MECCA), one of the first frameworks to support FIPA.

He is currently leading a variety of projects using agent technology with Siemens' operating divisions. In particular, he is Project Coordinator of Germany's most advanced traffic telematics project, the federally-funded Personal Travel Assistance (MOTIV-PTA) project, comprised of 9 leading German car manufacturers and IT companies. He has given a number of tutorials, short courses and invited talks in the areas of DAI and Multi-Agent Systems.

suguri1.jpg (52328 bytes)

Hiroki Suguri is a Director of Software Engineering Division of Comtec Corporation. He serves FIPA as a director since 1998. He is a vice president of Intelligent Agent Society of Japan. At Comtec, he has led one of the first developments of FIPA-compliant agent system. At FIPA, he is a chairperson of Interop Technical Committee that is responsible for interoperability trials, field trials, application development, verification of the specifications and maintenance of the specifications. Before joining Comtec, he worked for Data General Corporation in Westborough, Massachusetts, where he was involved with various software internationalization and localization projects.


You are visitor # since 20 August 1999