A Case-Based Approach to Realising Ambient Intelligence among Agents

This thesis presents a multi-agent based architecture and implementation to realising ambient intelligence. The architecture separates the abilities of an ambient intelligent system into three main areas: the ability to perceive the world which the system inhabits, the ability to reason about the state of the world by being context aware, and the ability to behave intelligently in occurring situations by being context sensitive. The aim of this work is to consider the notion of ambient intelligence from a holistic perspective, where all the aforementioned abilities are accounted for. Furthermore, this work argues that the ability to classify situations by being context aware is closely coupled with being knowledgeable about the world. Thus, achieving context awareness is approached from a knowledge intensive perspective, in particular through the episodic perspective of case-based reasoning. Case-based reasoning is utilised as the method of choice to assess and classify situations, thereby identifying the goal of the system in particular situations. This work argues that as an ambient intelligent system is situated in occurring situations involving human users, the ability to assess and classify situations is closely linked with understanding ongoing human activities. The knowledge model developed as part of this work combines the traditional pervasive computing perspective on context, with the aforementioned episodic perspective, and a socio-technical perspective on activities. The work presented in this thesis builds on both theoretical and empirical work. The underlying ideas were based on the literature and experience gained through the AmbieSense EU project. The main contributions of this work are a coherent conceptual architecture for realising ambient intelligences, a knowledge intensive method for modelling and representing context as an integrated part of a system’s knowledge model, and applying a small foot-print knowledge intensive case-base reasoning agent to assess and classify ongoing situations.