CONTEXT-AWARE AGENTS
The 6Ws Architecture
Juan Carlos Augusto
School of Computing and Mathematics and Computer Science Research Institute, University of Ulster, U.K.
John O’Donoghue
Business Information Systems, University College, Cork, Ireland
Keywords: Multi-Agent Systems, Agents architectures, Ambient Intelligence, Context-Awareness.
Abstract: Software agents have been designed and implemented to function within limited context-aware capabilities.
For an agent to function correctly and efficiently it should contain sufficient knowledge and reasoning
resources enabling them to process large quantities of implicit information conveyed through an explicit
description. Presented in this position paper is an introduction of the 6Ws agent-based architecture which
encompasses key reasoning capabilities which are not adequately supported by existing BDI frameworks
but have been recognized as highly relevant for the development of Ambient Intelligent systems.
1 INTRODUCTION
Ambient Intelligence, AmI, refers to ‘a digital
environment that proactively, but sensibly, supports
people in their daily lives’ (Augusto 2007). Other
terms such as Ubiquitous Computing (Weiser, 1991)
or Smart Environments (Cook and Das, 2005) are
used with similar connotations. Supporting people in
their daily lives means, for example, making an
environment safer, more comfortable and more
energy efficient.
What all these areas have in common is that they
are intended to operate in a specific environment.
Examples of such environments are homes,
classrooms, cars, offices. For these systems to be
truly ‘smart’ they need to perceive the interaction
with the end users and need to know as much as
possible about the environment itself (objects, user
preferences, latest changes, etc).
Software agents are well established as one of
the key enablers in delivering intelligent systems.
These agent based architectures have been
successful in delivering general purpose intelligent
systems, however the complexity of the
environments in AmI requires a more precise focus,
such as the capability to naturally specify context-
awareness features. Different concepts have gained
recognized relevance in building AmI systems:
Who: the identification of a user and the role that
user plays within the system.
Where: the tracking of the location where a user
or an object is geographically located at each
moment during the system operation.
When: the association of activities with time is
fundamental to build a realistic picture of a system’s
dynamic.
What: the recognition of activities and tasks
users are performing is fundamental in order to
provide appropriate help if required. The multiplicity
of possible scenarios that can follow an action
makes this very difficult. Spatial and temporal
awareness help to achieve task awareness.
Why: the capability to infer and understand
intentions and goals behind activities is one of the
hardest challenges in the area but with no doubt a
fundamental one which allows the system to
anticipate needs and serve users in a sensible way.
hoW: the alternative ways to achieve things in
the given environment. An architecture which
supports the previous five concepts will in turn
provide the hoW with sufficient supporting
information to make the correct decisions in a timely
manner.
Some of the required features as outlined above
may ultimately at present be obtained through
complicated, ad-hoc programming or through a
591
Augusto J. and O’Donoghue J. (2009).
CONTEXT-AWARE AGENTS - The 6Ws Architecture.
In Proceedings of the International Conference on Agents and Artificial Intelligence, pages 591-594
DOI: 10.5220/0001745305910594
Copyright
c
 SciTePress

BDI
6Ws
Limited
Attention
collection of technologies but developers in the area,
and ultimately users through higher quality products,
will benefit of having a specialized context-aware
developing framework.
The purpose of this paper is to propose an
architecture which relates to the current state of the
art and highlights the importance of concepts which
are fundamental for the facilitation of Ambient
Intelligence in Smart Environments.
2 AGENTS: CURRENT STATE OF
THE ART
The software agent-based paradigm is considered
highly suitable for constructing modular software
systems capable of operating in dynamic,
unpredictable environments (Koch, 2004). They
provide an established framework for analysing,
specifying, and implementing complex software
systems and can act as intelligent aids to users in
delivering advanced pervasive services. Agents
possess adept decision-making capabilities which
make them ideal for operating within dynamic
distributed networks.
In the context of software engineering, an agent
can be defined as: “An entity within a computer
system environment that is capable of flexible,
autonomous actions with the aim of complying with
its design objectives”.
Within a context-aware environment the
dynamic interaction between objects need to be fully
supported by intelligent software architectures. Such
environments generate vast arrays of implicit and
explicit information. An effective software
architecture will maximize this data source to deliver
the required services in a timely manner. In essence
the agent paradigm may be seen to anticipate the
needs of user and act on their behalf.
The agent paradigm has been applied to a
number of rich context-aware environments
including: as part of a real-time healthcare decision
support system in the deployment of an ambulance
services, (O’Donoghue, 2005) in the collection,
correlation and dissemination of real-time body area
network (BAN) sensor readings.
The agent paradigm has promised a great deal
and has delivered in certain aspects. A large
selection of software agent architectures have been
developed with a number of inherent design
philosophies with BDI (Beliefs, Desires and
Intentions) asserting itself and morphing as the de
facto approach.
With the BDI model Beliefs represent the
informational state of the agent, Desires represent
it’s motivational state i.e. it’s overall objective and
Intentions represent the deliberative state of the
agent, what the agent has chosen to execute.
In comparison AgentSpeak(L) (Rao, 1996) is
based on a strong theoretical foundation of logic
programming enabling it to explicitly define the role
of a particular agent in a declarative way. One of the
best recognised implementations of an
AgentSpeak(L) related agent architecture is a Java
interpreter called Jason. The interpreter was
developed to help integrate it within a variety of
applications. From a developers perspective the
AgentSpeack(L) provides an advantage over JadeX
in that it enables the designers to pay greater
attention to the overall AmI aspects without having
to sacrifice overall flow of reasoning in allowing for
the amalgamation of external components.
The Belief Desire Intention (BDI) software
model is an abstract designed primarily for software
agents. It is capable of separating the activity of
selecting a plan from the execution of a selected
plan. The BDI model has its limitations and is not
ideally suited for certain types of behaviour. “There
is a need for agent systems that can scale to real
world applications, yet retain the clean semantic
underpinning of more formal agent frameworks”
(Morley et al., 2004). In relation to ambient
intelligence and spatial and temporal reasoning BDI
models do have their limitations, for example with
BDI one can equate “Desires”-to-“What” and
“Intentions”-to-“Why” (c.f. figure 1).
Figure 1: BDI vis-a-vis 6Ws Scope.
Although the “hoW” element is not highlighted
in the BDI philosophy, it is usually present in the
way of a plan base available to the agent. However
the “Who”, “Where” and “When” elements are not
so faithfully represented in a BDI based architecture,
still they are essential elements in ambient intelligent
systems.
For any software agent architecture to fulfil its
true potential it needs to have the capability to relate
or understand its real world environment. One such
Who
Where
When
What
Why
hoW
ICAART 2009 - International Conference on Agents and Artificial Intelligence
592

approach is known as the Semantic Web. The
semantic web may be viewed as a web of
information which is structured and linked up in
such a way enabling other applications view and
understand that data i.e. providing a foundation for
agents to communicate and understand one another.
3 ENVIRONMENTS WHICH
DEMAND
CONTEXT-AWARENESS
Probably the most well known of such an
environment is a “Smart Home”. By Smart Home
here, we understand a house equipped to bring
advanced services to its users. There is a plethora of
sensing/acting technology, ranging from those that
stand alone (e.g., smoke or movement detectors), to
those fitted within other objects (e.g., a microwave
or a bed), to those that can be worn (e.g., shirts that
monitor heart beat). For example, in the case of
people at early stages of senile dementia (the most
frequent case being elderly people suffering from
Alzheimer’s disease) the system can be tailored to
minimize risks and ensure appropriate care at critical
times by monitoring activities, diagnosing
interesting situations and advising the carer. There
are already many ongoing academic research
projects with well established Smart Homes research
labs in this area, for example Domus (Pigot et al.,
2002), iDorm (Callaghan et al., 2001) MavHome
(Cook, 2006), and Gator Tech Smart Home (Helal ,
2005).
4 THE 6WS CONCEPT
This section focuses on the initial work to design
and develop a 6W agent based architecture to help
recognise and integrate all 6 aspects (Who, Where,
What, When, Why and hoW) which are relevant for
the implementation of AmI. At a logical level, a
representation of the 6Ws architecture is compared
against the BDI model c.f. figure 2.
Whilst “context” has been defined in many ways
several years of research in Ambient Intelligence
have highlighted the importance of certain elements
in the success of building systems within this area.
A consensus in this area is that systems with
Ambient Intelligence should be built as human-
centric, systems should serve humans and not vice-
versa. Systems should be able to learn about the
needs and preferences (compare “the user need to
increase insulin intake to keep glucose at the right
level” with “I prefer to minimize the number of
insulin intakes”) of the users they are supposed to
serve and, if necessary and feasible, hold updated
profiles of them to ensure they can accomplish their
service in the best possible way.
Figure 2: The 6Ws attention to key ambient intelligent
elements i.e. profile, situation and temporal. c.f. figure 3 in
relation to the AgentSpeak(L) semantic model.
The rationale for the inclusion of the Who
component is that an important part of the
meaningful context the agent should know is the
needs and preferences of the potential users. Fig. 3
provides a depiction on our modified AgentSpeak(L)
architecture. [1] highlights the Who component
(which emphasise the user-centred characted of the
system) as an important component of the Belief
Base, [2] and [3] embeds the Where (spatial
conditions) and When (temporal conditions)
elements within the meaningful events which can
describe triggering situations, [4] includes the Why
dimension which highlights the desires of the agent
(this includes paying attention to the needs and
preferences related to users as specified in the Who
dimension), [5] highlighting the What component
through the specification of intentions, and [6] refers
to the hoW component as the plans represent the
ways the agent can achieve the goals.
5 CONCLUSIONS
Presented is the concept of developing an agent
based context-aware architecture with 6 elements
which are key to the development of Ambient
Intelligence system with a rational for the Who,
Where, When, What, Why and hoW elements. As an
initial step the 6Ws approach was compared against
the well established BDI model. We have
highlighted that the BDI model still contains a
number of weaknesses as a framework for AmI e.g.
user-centeredness, spatial and temporal reasoning
which are within Ambient Intelligence systems.
Our proposed framework although following a
route closer to AgentSpeak for the implementation
1 Who Profile (including
N
eeds/Preferences)
Beliefs
2 Where Spatial conditions
3 When Temporal conditions
4 Why General aims Desires
5 What Specific goals
Intentions
6 hoW Selection of plans
CONTEXT-AWARE AGENTS - The 6Ws Architecture
593

Figure 3: The AgentSpeak(L) semantic model with the 6Ws modifications to include key AmI elements.
of this architecture the idea is actually independent
of the final implementation of choice currently under
development. We are currently given further steps
directed to have an implemented framework which
supports the 6Ws extended BDI architecture.
REFERENCES
Augusto J.C., 2007. Ambient Intelligence: The Confluence
of Ubiquitous/Pervasive Computing and Artificial
Intelligence, pages 213–234. Intelligent Computing
Everywhere. Springer London.
Callaghan V., Clarke G., Colley M., and Hagras H., 2001.
A Soft-Computing DAI Architecture for Intelligent
Buildings, Studies in Fuzziness and Soft Computing,
V. 75, Soft Computing Agents: New Trends for
Designing Autonomous Systems. pp. 221, Springer.
Cook D., 2006, Health Monitoring and Assistance to
Support Aging in Place. JUCS, 12(1):15–29, 2006.
Cook D.J. and Das S.K., 2005. Smart Environments:
Technology, Protocols and Applications. Wiley-
Interscience.
Koch F. and Rahwan I., 2004. Classification of Agent-
Based Mobile Assistants, In Proc. of the Autonomous
Agents and Multi-Agent Systems Workshop on Agents
for Ubiquitous Computing, pp. 7-38.
Helal A., Mann W., Elzabadani H., King J., Kaddourah
Y., and Jansen E., 2005. Gator Tech Smart House: A
Programmable Pervasive Space. IEEE Computer
magazine, pages 64–74, March 2005.
Morley D., and Myers K., 2004. The SPARK Agent
Framework. In Proceedings of the third International
joint conference on Autonomous Agents and Multi
Agent Systems (AAMAS), pp. 712-719.
O'Donoghue J. and Herbert J., 2005. Data Management
System: A Context Aware Architecture for Pervasive
Patient Monitoring, Proceedings of the 3rd
International Conference on Smart Homes and Health
Telematic, pp. 159-166.
Pigot H., Mayers A., Giroux S., Lefebvre B., Noury N.
Rialle V.,2002. Smart House for Frail and Cognitive
Impaired Elders.
Rao A.S., 1996. AgentSpeak(L): BDI Agents Speak Out in
a Logical Computable Language, In Proceedings of
the 7th European Workshop on Modelling
Autonomous Agents in a Multi-Agent World, vol. 1038
of LNCS, pp. 42-55. Springer.
Weiser M., 1991. The Computer for the 21st Century.
Scientific American, 265(3):94–104, 1991.
ICAART 2009 - International Conference on Agents and Artificial Intelligence
594