Smart Home Control Based on Behavioural Profiles

Abstract:
Automation of homes and buildings is characterized by a high casuistry and a strong interaction with users. Here, concepts like ubiquity and cooperativeness acquire an undeniable importance. From a top-down, holistic perspective, this thesis states a methodology for the design of adaptive Smart Homes where users' habits are considered as context elements able to join home applications, services and technologies in a sound overall performance. Furthermore, the focus on users' behaviours allows the exploration of persuasive, proactive designs oriented to improve system adaptiveness. The work relies on the literature as a starting point for the establishment of definitions and taxonomy for behavioural profiles. It yields to the design of an overall smart home control concept based on habit profiles, describing the integration within current home automation networks, and also considering synergies with regard to wider scopes on the way to smart cities (e.g. building calculations, urban planning, etc.). The proposal gains soundness by the description of a set of profile-based applications that cover main functionalities of automated homes. Among them, the use cases "air quality" and "thermal confort" are selected for a deep analysis, and algorithms and structures are developed and described in detail. The second part of the study focuses on the accurate elaboration of profile-based controllers. It requires dealing with complexities related to users' behaviour modeling, habit abstraction and context interpretation. Such demanding tasks are faced by clustering processes grounded by Soft Computing techniques, advanced solutions that entail a careful design and parameterisation in a close relationship with the peculiarities of the intended applications.
The diverse proposals and discussions offered throughout the dissertation are evaluated by means of mathematical analysis and comparisons carried out in a test-bed simulated environment, culminating in a set of generalisations and rules of application. In addition, the Exclusive Self-Organising Map algorithm (XSOM) has been developed as a searching technique for pattern discovery and context inference that perfectly fit the requirements of the field under study.
In this dissertation, the holistic habit-based smart home consolidates as a promising approach mainly due to the following features: it enhances the control of isolated applications; it grounds global, coordinated management, improving thereby the overall performance; finally, it allows the development of high user-sensitive environments.