IIDS Research
Research themes:
- Middleware
- Services
- Applications
- Legal Aspects
The number of agents and resources in Internet-scale applications can
grow to thousands, ten thousands, hundred thousands, and more. Agent
platforms need to be able to support such large scale multi-agent
systems. They need to provide efficient and effective means for
life-cycle management, communication, resource management, security,
fault-tolerance, etc. These basic principles are topics addressed
within the design and implementation of the AgentScape platform.
The AgentScape Operating System consists of two layers: the
AgentScape Operating System kernel and the AgentScape Operating System
middleware. The AgentScape Operating System kernel provides the basic
mechanisms for higher level functionality and policies for agent
platforms, e.g., resource access and management, fault-tolerance,
security policies, etc.
Management of agent platforms in an Internet-scale, open system
setting is a complex task. A management system should take both agent
and resource specific requirements into account. The scale of
deployed agents and resources, and the heterogeneous requirements,
dictates a decentralized approach and design of the management system
for the agent platform. The decentralized approach of system
management is based on self-management, which includes
self-configuration, self-healing, self-optimization, and
self-protection.
The Automated Middleware Testing environment provides a suite of
tests and a graphical user interface to evaluate and assess the
correct implementation of functionality in the agent platform. The
test suite includes basic tests for agent life-cycle management,
communication, and migration. Besides the basic tests, more complex
tests are included that mimic typical behavior of multi-agent system
applications.
Agentscape middleware and services provide support for a wide variety
of multi-agent system applications. However, AgentScape does not
provide a framework for multi-agent system application development.
Mansion has been designed for this purpose. Mansion introduces and
supports a new paradigm. Applications are designed as worlds with
rooms in which agents can roam, accessing (replicated) objects and
communicating with other agents located in the rooms. Mansion is
built on top of AgentScape and provides additional functionality
supporting this paradigm.
- Directory Services
- Software Configuration
Software configuration services and
directory services are the two types of services currently addressed. Both types of services are not developed
specifically for AgentScape: they are intended to be of use to other
supportive environments for distributed applications. In essence, a
service offers an interface and is usually associated with a specific
location.
- Software configuration services are used to automatically create and
maintain agents and services. The IIDS group researches
principles
of software configuration, and applications to both
agent-configuration and (web-)service configuration. An
associated project, in which software configuration is combined with
migration, yielding cross-platform heterogeneous
generative migration, is used to increase interoperability between
different agent platforms, including Jade and AgentScape.
- Directory services are used to look up agents and services with
specific characteristics. As centralistic approaches are unrealistic
in highly dynamic large-scale distributed agent systems, a decentralised solution is
researched, which uses location services to store and retrieve
agents' and services' locations, and decentralised directory
services to store and retrieve annotations of agents and services.
Within the AgentScape research programme, distributed applications are
the proof of the pudding: they determine the requirements for
AgentScape's middleware and services and also the means with which the
former can be tested. Three major application areas are currently
explored: distributed design, agent based market places, and
agent based social simulation.
- Distributed design involves single, or multiple collaborative,
designer agents in interaction with their environment to solve
(complex) design problems. This research requires an understanding
of design, but also of multi-agent co-ordination, collaboration and
trust; research areas which are open to much discussion. This
research builds on our research on design since 1992 and agents
since 1995.
- Agent based market places provide a playground for
highly-interactive multi-agent applications, in which strategies and
interoperability play an important role. Negotiation requires agents
to adapt their strategies to those of other agents (agents with
which they negotiate). An electronic market place is used in the
assignments given during a graduate seminar provided by the IIDS group. Changing languages and
protocols for information sharing and negotiation show the need for,
and application of, automated software configuration services.
- Agent based social simulation involves large numbers of
interactive and mobile agents participating in a simulation to study
phenomena of traditionally social scientific interest, such as
cooperation, coordination, organizational behavior, social dynamics,
group and coalition formation, and the evolution of conventions and
norms.
Agent characteristics -- such as: autonomy, pro-activity, reasoning, learning, co-operation,
negotiation, social and physical manifestation -- are characteristics taken from humans.
Notions such as anonimity and privacy will be redefined in the 'digital world'. In these projects
the research areas of Computer Science, Artificial Intelligence, and Law are combined to analyse
legal possibilities and limitations of agent technology. The aim of these projects is to provide
guidelines for both AI-researchers and Legal-researchers.
- The ALIAS project
analyses legal and technical implications of
design choices underlying software agent applications. The complex
interplay between technical progress and legal advances at national
and international levels is studied in an interdisciplinary setting.
The advent of agent technology, in which mobile, autonomous agents
roam the Internet and interact with other agents on behalf of their
users, challenges legal assumptions (e.g., can agents legally sign
contracts) and is regulated by legal restrictions (e.g., are agents
and/or their human owners required to be traceable). The acceptance of
agent technology as a basis for large-scale, flexible, heterogeneous
distributed applications hinges on legal support.
-
The ACCESS project
focuses on whether, and to which extent, software agents can be used to maintain the integrity of information
in digital dossiers. This information is acquired from different, distributed sources, and used in the
semi-closed environment of the Courts in Amsterdam and Rotterdam.This project proposes a new
approach to information processing, supporting court officials to a better degree
than currently is the case, based on the availability of the digital dossier. This project focuses
on the automation of work processes related to the use of digital dossiers in the criminal courts.
Two of the areas of relevance are: (1) control and (2) navigation, adaptation and presentation.
The autonomous, pro-active nature of agents influences the division of responsibility and initiative
in the work processes involved. The integrity of the results is, however of utmost importance.
As with all technology navigation, adaptation and presentation, are the aspects which will
strongly influence the effect of its use.
