Agent-Based Computing : APPLICATION OPPORTUNITIES

1.6.1 Ambient Intelligence
The notion of ambient intelligence has largely arisen through the efforts of the European Commission in identifying challenges for European research and development in information society technologies. Aimed at seamless delivery of services and applications, it relies on three identified pillars of ubiquitous computing, ubiquitous communication, and intelligent user interfaces, yet it offers perhaps the strongest motivation for, and justification of, agent technologies. The ambient intelligence vision describes an environment of potentially thousands of embedded and mobile devices (or software artifacts) interacting to support user-centered goals and activity. This suggests a component-oriented view of the world in which the artifacts are independent and distributed. The consensus is that autonomy, distribution, adaptation, responsiveness, and so on, are the key characterizing features of these ambient intelligent artifacts, and in this sense they very strongly share the same characteristics as agents.

In particular, these ambient intelligence artifacts are likely to be function-specific (though possibly configurable to tasks) and will, of necessity, need to interact with numerous other ambient intelligence artifacts in the environment around them in order to achieve their goals. Interactions will take place between pairs of artifacts (in one-to-one cooperation or competition), between groups of artifacts (in reaching consensus decisions), and between artifacts and the infrastructure resources that comprise their environments (such as large-scale information repositories or other supporting resources, possibly through agent encapsulation). Interactions like these enable the establishment of electronic institutions or virtual organizations, in which groups of agents come together to form coherent groups able to achieve some overarching goals.

1.6.2 Grid Computing
The high-performance computing infrastructure, known as the Grid, for supporting large-scale distributed scientific endeavor has recently gained heightened and sustained interest from several communities, as a means of developing e-science applications such as those demanded by the Large Hadron Collider facility at CERN (the European Organization for Nuclear Research), engineering design optimization, and combinatorial chemistry. Yet it also provides a computing infrastructure for supporting more general applications that involve large-scale information handling, knowledge management, and service provision.

It is natural to view large systems in terms of the services they offer, and consequently in terms of the entities providing or consuming services. Grid applications, in which typically many services may be involved, spread over a geographically distributed environment, which new services join and existing ones leave, thus very strongly suggest the use of agent-based computing. In this view, agents act on behalf of service owners, managing access to services, and ensuring that contracts are fulfilled. They also act on behalf of service consumers, locating services, agreeing to contracts, and receiving and presenting results. Just as in the ambient intelligence vision, agents will be required to engage in interactions, to negotiate, and to make proactive run-time decisions while responding to changing circumstances. In particular, agents will need to collaborate and to form coalitions of agents with different capabilities in support of new virtual organizations. Such virtual organizations have been identified as the tool with which to unwrap the power of the Grid.

Initially geared towards high-performance computing, Grid computing is now being recognized as the future model for service-oriented environments, within and across enterprises. The impact will be larger than just virtual organizations—a global company is much like such a virtual organization and will require similar technology.

1.6.3 Electronic Business
To date agents have been used in the first stages of e-commerce, product and merchant discovery, and brokering. The next step will involve moving into real trading, negotiating deals, and making purchases. However, it can be argued that the real impact of electronic commerce will be on a dramatic change in the supply chain. If a consumer can contact directly the producer instead of a reseller it might produce an increase in efficiency of the overall supply chain. These changes in the supply chains will permit new markets to appear, old markets to change, and the participation of new players. These observations raise some broader questions about e-commerce in general, and the speeding-up effects of agents in particular. Consumers who are excluded from the e-commerce loop may find their prices and choices becoming worse.

In the short term, travel agencies and retailing will be the primary business-to-consumer application domains using agent technology in e-commerce. One of the current efforts aimed at driving this forward can be seen in the Trading Agents Competition (TAC), which offered a sophisticated problem domain of multiple auctions for agents to compile travel packages for customers. Such initiatives can highlight the potential of agent technology for a wider audience, while at the same time contributing to the more rapid development of the field in a specific application and problem domain. Here, one interesting segment is supply chain management for virtual and transnational enterprises. On the other hand, it can be foreseen that agent technology in this market will enable small and medium enterprises to collaborate and form coalitions in much more flexible ways, almost regardless of geographic location.

In the longer term, full supply chain integration is the aim. According to a Pricewaterhouse-Coopers report, there were more than 1,000 public e-markets and around 30,000 private exchanges at the beginning of 2001. Although the baseline domains exist, the lack of standards and uniformity of these platforms constrains what can currently be achieved, but offers a real challenge and opportunity for deployment of agent systems over the next 5 to 10 years.

1.6.4 Simulation
Multiagent systems provide a natural basis for training of decision-makers in complex decision-making domains. For example, defense simulations using multiagent systems can enable military planners, strategic defense staff, and even operational staff to gain experience of complex military operations through simulations and war games. These simulated experiences are obtained instead of, or in addition to, experiences gained in actual military operations. Similarly, decision makers in other complex and dynamic environments can gain valuable experience through exercises that simulate their real-world domain using multiagent systems. Applications include marketplaces subject to rapid change, such as telecommunications markets undergoing deregulation, and markets for fast-moving consumer goods, such as breakfast cereals, where consumer tastes and competitor activities can lead to market turbulence. In these applications, as for those in defense, multiagent systems may simulate over a few hours the dynamics of an actual market that could occur over several years, and so give trainee decision makers rapid exposure to many diverse experiences. In addition, as the military example reveals, the decision maker is allowed to learn through his or her mistakes without creating real-world consequences.

Social simulation is somewhat unusual in that it does not require many of the challenges listed earlier to be addressed for it to succeed in the time scales considered in this report. Since simulations are by their nature closed (even though they may model open systems), they are almost immediately enabled. However, there are many open issues to be resolved before agent-based simulation models can be applied more widely to public policy domains. For example, there is as yet no general understanding of what constitutes "good" performance by a multiagent system, except perhaps in some domains. There is no guarantee, for example, that an agent society in which different species of agents coevolve in the course of their interactions with one another will progress in any sense; later generations of a species may be less fit than earlier generations of that same species when pitted against earlier generations of their competitor species. In such a case, at which timepoint should the simulation be terminated? Different termination points may lead to different assessments of system performance and different recommendations to policy makers. Indeed, the question of performance assessment of multiagent systems is part of a larger, mostly open, question of performance assessment of decision support systems in general.

Multiagent systems as social simulations are also of increasing importance in entertainment applications. These applications range from single (human)-player computer games to multiplayer games, where the other players may be both humans and agents. Potential applications also exist in other interactive media, such as interactive movies, television, and even books, where viewers and readers may have their own avatar participate in the story and may interact with fictional characters directly. The film industry has already notched up several successes in this area, perhaps the most notable recent example being the film, The Lord of the Rings: The Two Towers. This film achieved visually impressive battle scenes by using agent technology to model individual characters, with the overall film sequence emerging from their interactions.