Context Awareness in the Internet of Things and its Applications

In the last decade the role of context awareness, traditionally focused on human to machine interaction, has broadened its perspectives to the machine to machine paradigma. The main goal of this dissertation is both to understand how to apply context awareness to situations , perceived by smart devices, and to conceptually stretch context modeling to a dynamic contextualization in human to machine interactions. The roots of the Internet of Things paradigma reside in the efforts reached in the Wireless Sensor Networks technology, mostly in data aggregation and in energy saving, and the adoption of multi-agent modeling has gained context awareness application to machine to machine interaction. A well defined methodology, previously applied to human to machine interaction, now can be adopted for smart devices, that behave like humans. Another evidence of the emergence of the Internet of Things technology nowadays comes from the everyday life experience. The Internet of Things is the key for the practical implementation of innovative software systems for the ubiquitous computing. Thanks to all these technologies, Smart devices (like the Nest Thermostat or the Apple Watch) are currently more and more integrated among them and they are becoming invisible servants for final users. As a proof of this new technological era, we can think about how the usage of the Siri tool has become an automated and unconscious mechanism in looking for a telephone number, in reaching a specific destination, or in driving home heating, to understand the potentialities of merging context awareness to the Internet of Things in a convergent and ubiquitous platform. The enormous amount of smart devices, currently deployed in the world, have also to deal with an easy knowledge representation of sensed context, in order to provide new mechanisms to automate daily tasks, understanding the behavior of end-users within the surrounding environment. On the other hand, the rapid growing number of smart devices deployed has a drawback in the future proliferation of high level context models, possibly coupled to lower context levels. What emerges from the current study is the necessity to easy the management of multiple contexts, to be used by upper level applications. The dynamic contextualization solves this kind of issues, distinguishing from the total amount of features, captured from the surrounding environment, and the context model that is closely related to the issue to be solved. Deep profiling on context aware usage enhances the development of context aware services, that can simply use an abstraction layer to properly manage underlying context models. What can be deduced is that the customization of context aware services to the user is a key process to narrow the gap between smart devices and their daily usage. In this dissertation, the definition of high level scenarios have been determined by applying decision trees, for their huge potentialities expressed in dynamic context extraction. Applications of these concepts were used in developing management systems, addressed to an audience of experienced surgeons in breast cancer, covering surgical suggestions. The formal analysis of multiple datasets (related to the diagnosis of breast cancer), using interactive and navigable decision trees, showed the enormous potentialities of the system, both in knowledge representation (and its spreading), and in the identification of the context, considering the related decision support system mechanisms. The conclusion of the research activity considers the emergence of context awareness in a future world, more and more full of smart devices connected among them, as an adaptive paradigm, for intra device optimizations and for final users application level benefits.