Dr. Aitor Almeida

Intelligent Environments and context-aware systems aim to improve users’ everyday life by taking care of daily activities while being completely invisible. In order to achieve this, three objectives must be accomplished: a) the reactions to the context changes have to be accurate, b) those reactions have to be timely and c) computers have to disappear into the environment. These objectives can be difficult to achieve at times. In order to achieve sensible and accurate reactions to changes in the environment, the model of the context must be as close to reality as possible. This model must be as expressive as possible to achieve a more precise inference over the context and thus to better cater to the needs of the users. On the other hand, achieving both timely reactions and environment integrated computational capabilities can be conflicting at times. Integrating the computational capabilities into the environment often results in using more limited devices, which in turn results in higher inference times. This dissertation aims to tackle these problems. In order to create a better context model, a new context reasoning mechanism that combines semantic inference with uncertainty and vagueness will be presented. This mechanism will allow to better model the environment, achieving more expressivity than traditional context models. The proposed mechanism also includes a data fusion algorithm that integrates the measures taken from different devices to create a unified vision of the environment. To provide timely reactions and help integrating the computational capabilities in the environment, a distributed inference architecture will be described. This architecture will allow splitting the global inference task into smaller inference sub-units, resulting in a faster, lighter, more efficient and more robust inference.