Intelligent Agents for Automatic Service Composition in Ambient Intelligence

The Ubiquitous Computing concept was first defined by Mark Weiser in (Weiser, 1995), referring to a new computing era where electronic devices merge with the background, becoming invisible, in such a way that people could make use of those devices in an unconsciously way, focusing just on their needs and not in the interaction. One decade later, the IST Advisory Group first states the concept of Ambient Intelligence (Ducatel et al., 2001), which lying on the ubiquitous computing paradigm, refers to those environments where people are surrounded by all kind of intelligent intuitive devices, capable of recognising and responding to their changing needs. In these contexts, people perceive the surrounding as a service provider that satisfies their needs or inquiries in a seamless, unobtrusive, and invisible way. These computing paradigms set a frame of reference, characterised by being mainly concentrated on releasing mechanisms that gather information about users, match behavioural patterns, or predict user actions, requirements and needs (Costa et al., 2007) (Cugola & Picco, 2006) (Issarny et al., 2005) (Prete & Capra, 2008). Nevertheless, the Ambient Intelligence paradigm is meant to consider users as constituent parts of the context, although in most solutions presented to date, users are considered in isolation. In this regard, extending the user-centered view, in order to encompass the context services and purposes, arises as key requirements for systems in Ambient Intelligence. It soon becomes apparent the need for a multidisciplinary approach capable of addressing all the emerging challenges. One of these fields is concerned with the communication support. The heterogeneity of the context devices, as well as their dinamism, impose high demands upon the middleware platform, that it is now responsible for abstracting the technological peculiarities. It is then possible to provide a common and well-known set of communication interfaces. These interfaces are described in terms of a semantic model, that can be easily shared and translated into different languages, so that in can be used by the rest of involved technologies (intelligent agents and reasoning engine). Finally, and probably the most important part of the provided solution, refers to the context-awareness, in charge of understanding the context. This requires some approach that resembles human behaviour in regard to its ability to deal with information of an imprecise nature, ambiguous, and of a questionable retrievability, but also capable of making decisions based on this partial information. To this 2