Object (B)logging: a Semantic-Based Self-Description for Things Networks in Pervasive Contexts

This thesis proposes Object (b)logging, a novel general approach for a Semantic Web of Things, based on an evolution of conventional Web of Things paradigms and introducing ubiquitous Knowledge Base KB models in order to associate semantic annotations to real-world objects and events. Object (b)logging represents the capability of an object to describe itself and its context in a fully automated fashion starting from raw environmental data collected by sensors. The overall goal is to define a knowledge-based framework for high-level information representation, knowledge discovery, allotment and sharing in distributed scenarios populated by smart objects. Smart object is an intelligent agent equipped with embedded sensors, actuators, communication interfaces, computation and storage. Several heterogeneous micro-devices cooperate to connote and modify appropriately the state of the surrounding environment. By leveraging the integration of standard machine learning techniques with non-standard semantic-based reasoning services, the dissertation defines software architectures and methods to enable efficient automated context annotation on resource-constrained mobile computing devices and to progressively improve produced descriptions during the object's lifetime. Throughout the objects lifetime, the acquired knowledge is exposed to the outside world as in a blog: to achieve this, the proposal includes a layered architecture built on a publish/subscribe Message Oriented Middleware. A novel collaborative and distributed information sharing approach is enabled in pervasive computing scenarios featured by volatile nodes interacting in an opportunistic fashion. It is based on the ubiquitous KB (u-KB) paradigm, which allows to manage in a decentralized way knowledge scattered on several nodes within a network. Semantic matchmaking is exploited to support dynamic and flexible knowledge discovery. The various elements of the framework were implemented in suitable prototypical testbeds and experimental analysis was carried out with reference to selected case studies. Results indicate the feasibility and usefulness of the envisioned approach.