A Pervasive System Architecture that supports Adaptation using Agents and Ontologies Achilles Kameas, Christos Goumopoulos The Hellenic Open University and DAISy research unit Patras, Hellas {kameas, goumop}@cti.gr Hani Hagras, Michael Gardner University of Essex Colchester CO4 3SQ, United Kingdom {hani, mgardner}@essex.ac.uk Tobias Heinroth, Wolfgang Minker Ulm University Ulm, Germany {tobias.heinroth, wolfgang.minker}@uni-ulm.de Apostolos Meliones, Dimitris Economou inAccess Networks S.A Athens, Greece {meliones, decon}@inaccessnetworks.com Yacine Bellik, Gaëtan Pruvost LIMSI – CNRS The National Center for Scientific Research Orsay, France {yacine.bellik, gaetan.pruvost}@limsi.fr Abstract—In the context of the EU funded R&D project ATRACO we are developing a conceptual framework and a system architecture that will support the realization of adap- tive and trusted ambient intelligent systems. Our approach is based on a number of well established engineering principles, such as the distribution of control and the separation of service interfaces from the service implementation, adopting a SOA model combined with intelligent agents and ontologies. Agents support adaptive task realization and enhanced human- machine interaction while ontologies provide knowledge repre- sentation, management of heterogeneity, semantically rich resource discovery and adaptation. ATRACO systems are dy- namic compositions of distributed, loosely-coupled and highly cohesive components that operate in dynamic environments. Keywords-adaptation; ontologies; fuzzy logic; SOA; HCI I. INTRODUCTION Ambient Intelligence (AmI) is a paradigm that puts for- ward the criteria for the design of the next generation of ubiquitous computing environments [1]. In the AmI para- digm, intelligent computation will be invisibly embedded into our everyday environments through a pervasive trans- parent infrastructure (consisting of a multitude of sensors, actuators, processors and networks) which is capable of recognizing, responding and adapting to individuals in a seamless and unobtrusive way. Several approaches have been proposed for realizing Ambient Intelligent Environ- ments (AIE), e.g., in [2], [3]. In these approaches, AIEs con- tain a dynamically changing set of “smart” objects or de- vices (hereafter called “artifacts”), which are able to per- ceive the environment, act upon it, process and store data, manage their local state, communicate and exchange data. As a result, artifacts may have physical properties (if they are physical objects), offer a set of services and can com- municate with each other and the environment. In addition, the AIEs provide an infrastructure that supports services such as networking, communication, discovery, location and context estimation. These services are used by the artifacts that reside within this environment. The next step is the design and development of totally adaptive ubiquitous computing systems, able to consistently operate in heterogeneous constantly changing AIEs. This Next Generation of AIEs (NGAIEs) will still be populated with numerous devices and have multiple occupants, but will inherently exhibit increasingly intelligent behaviour, provide optimized resource usage and support consistent functionality and human-centric operation. In our attempt to realize adaptive AIEs, we have come across various research challenges, such as heterogeneity of artifacts, system transparency, discovery & management of various artifacts, and autonomous behaviour of learning agents. The ATRACO (Adaptive and TRusted Ambient eCOlogies) approach presented in this paper addresses these challenges and produces specifications and concrete realiza- tion of adaptable AIEs. We have defined the concept of an Activity Sphere (AS), to be both the model and the realization of the set of information, knowledge, services and other resources re- quired to achieve an individual goal within an AIE. It is based on the notion of “bubble”, which has been used to describe a temporary defined space that can be used to limit the information coming into and leaving the digital domain [4], which itself constitutes a “digitization” of the definition of personal space described as a “soap bubble” [5]. Inspired by object-oriented approaches, an AS expands the bubble notion to contain not only the data and models, but also the associated processes and other resources that create, use or otherwise affect this data, leading to the specification of autonomous and coherent entities, which can adaptively execute on changing infrastructure. Following a similar perspective, we consider artifacts as having an internal part that encapsulates their internal struc- ture and functionality, and an external part that manifests their capabilities and can influence their environment. 2009 10th International Symposium on Pervasive Systems, Algorithms, and Networks 978-0-7695-3908-9/09 $26.00 © 2009 IEEE DOI 10.1109/I-SPAN.2009.81 148 Authorized licensed use limited to: Hellenic Open University. Downloaded on July 19,2010 at 10:03:24 UTC from IEEE Xplore. Restrictions apply.