Situation awareness in pervasive computing systems: reasoning, verification, prediction

Abstract

The paradigm of pervasive computing aims to integrate the computing technologies in a graceful and transparent manner, and make computing solutions available anywhere and at any time. Different aspects of pervasive computing, like smart homes, smart offices, social networks, micromarketing applications, PDAs are becoming a part of everyday life. Context can be defined as information that can be of possible interest to the system. Context often includes location, time, activity, surroundings among other attributes. One of the core features of pervasive computing systems is context awareness – the ability to use context to improve the performance of the system and make its behavior more intelligent. Situation awareness is related to context awareness, and can be viewed as the highest level of context generalization. Situations allow eliciting the most important information from context. For example, situations can correspond to locations of interest, actions and locomotion of the user, environmental conditions. The thesis proposes, justifies and evaluates situation modeling methods that allow covering broad range of real-life situations of interest and reasoning efficiently about situation relationships. The thesis also addresses and contributes to learning the situations out of unlabeled data. One of the main challenges of that approach is understanding the meaning of a newly acquired situation and assigning a proper label to it. This thesis proposes methods to infer situations from unlabeled context history, as well as methods to assign proper labels to the inferred situations. This thesis proposes and evaluates novel methods for formal verification of context and situation models. Proposed formal verification significantly reduces misinterpretation and misdetection errors in situation aware systems. The proper use of verification can help building more reliable and dependable pervasive computing systems and avoid the inconsistent context awareness and situation awareness results. The thesis also proposes a set of context prediction and situation prediction methods on top of enhanced situation awareness mechanisms. Being aware of the future situations enables a pervasive computing system to choose the most efficient strategies to achieve its stated objectives and therefore a timely response to the upcoming situation can be provided. In order to become efficient, situation prediction should be complemented with proper acting on prediction results, i.e. proactive adaptation. This thesis proposes proactive adaptation solutions based on reinforcement learning techniques, in contrast to the majority of current approaches that solve situation prediction and proactive adaptation problems sequentially. This thesis contributes to situation awareness field and addresses multiple aspects of situation awareness. The proposed methods were implemented as parts of ECSTRA (Enhanced Context Spaces Theory-based Reasoning Architecture) framework. ECSTRA framework has proven to be efficient and feasible solution for real life pervasive computing systems. Submitted in partial fulfillment of the requirements for the Doctor of Philosophy (Dual Award) (Lulea University of Technology)