Easing the smart home a rule-based language and multi-agent structure for end user development in intelligent environments
- García Herranz, Manuel
- Pablo A. Haya Zuzendaria
- Xavier Alamán Roldán Zuzendaria
Defentsa unibertsitatea: Universidad Autónoma de Madrid
Fecha de defensa: 2009(e)ko urria-(a)k 29
- Julio Abascal González Presidentea
- Germán Montoro Manrique Idazkaria
- José Bravo Rodríguez Kidea
- Francisco Ballesteros Cámara Kidea
- Diego Gachet Kidea
- Tiao-Yang Cai Kidea
- Juan Carlos Augusto Kidea
Mota: Tesia
Laburpena
As computing, networking and sensoring technologies evolve, increasing their speed and lowering their size and cost, the number of computerized elements present in our daily lives is ingrowing steadily and fast both in number and diversity. How many objects with processors, sensors or actuators do we already cross by in an ordinary day? How much information is retrieved somehow during the day about our lives? How many combinations of that information and capabilities may result which bene¿t us? To whom of us exactly? Despite their evolving number and diversity, most computing enriched elements found in the environment are isolated from one another, created by di¿erent third party developers, strange to the end¿users lives and particular needs. This is, in turn, a problem of control. Problem of special relevance in personal environments which are free¿choice environments and strongly involved in families and individual self¿de¿nition. Starting from this position, the present thesis analyses the problem of putting the end¿users in the loop of deciding the behaviors of their own environments, analyzing, due to the heterogeneity of users, preferences and environments, the adequateness of expression and structuring mechanisms. What sort of language is best to deal with unskilled users with simple preferences while allowing to express highly complex behaviors? How can we deal with multiple inhabitants and domains of automation? What may happen when di¿erent users with con¿icting preferences share the same environment? How can we bene¿t from automatic learning while preserving predictability? What sort of intelligence is ¿t to co¿live with people in their personal spaces? In response to these questions, an Event Condition Action (ECA) rule language is presented as the underlying kernel language for application¿ independent control. This language is designed as an UI¿independent expression and explanation mechanism in Intelligent Environments, opening the door to predictable automatic learning. The naturalness of the language and its adequateness to deal with non¿programmers has been studied through an end¿user survey, measuring the resemblance of natural programming structures with those of the kernel language, while its complex expression capabilities have been compared with state of the art event composition algebras. In addition, a multi¿agent structure is presented to organize and manage multi¿preference environments, allowing users to translate their natural hierarchies in their enriched environments, replicating the di¿erent degrees of complexity present in human organizations. The system presented in this work has been deployed and tested in various environments of di¿erent nature as well as in combination with other state of the art technologies. The particular challenges present in every environment, as well as the synergistic and wrapping capabilities of such a control mechanism extracted from our experience, are ¿nally exposed.