Cognitive Models and Computational Approaches for improving Situation Awareness Systems

Abstract

The world of Internet of Things is pervaded by complex environments with smart services available every time and everywhere. In such a context, a serious open issue is the capability of information systems to support adaptive and collaborative decision processes in perceiving and elaborating huge amounts of data. This requires the design and realization of novel socio-technical systems based on the “human-in-the-loop” paradigm. The presence of both humans and software in such systems demands for adequate levels of Situation Awareness (SA). To achieve and maintain proper levels of SA is a daunting task due to the intrinsic technical characteristics of systems and the limitations of human cognitive mechanisms. In the scientific literature, such issues hindering the SA formation process are defined as SA demons. The objective of this research is to contribute to the resolution of the SA demons by means of the identification of information processing paradigms for an original support to the SA and the definition of new theoretical and practical approaches based on cognitive models and computational techniques. The research work starts with an in-depth analysis and some preliminary verifications of methods, techniques, and systems of SA. A major outcome of this analysis is that there is only a limited use of the Granular Computing paradigm (GrC) in the SA field, despite the fact that SA and GrC share many concepts and principles. The research work continues with the definition of contributions and original results for the resolution of significant SA demons, exploiting some of the approaches identified in the analysis phase (i.e., ontologies, data mining, and GrC). The first contribution addresses the issues related to the bad perception of data by users. We propose a semantic approach for the quality-aware sensor data management which uses a data imputation technique based on association rule mining. The second contribution proposes an original ontological approach to situation management, namely the Adaptive Goal-driven Situation Management. The approach uses the ontological modeling of goals and situations and a mechanism that suggests the most relevant goals to the users at a given moment. Lastly, the adoption of the GrC paradigm allows the definition of a novel model for representing and reasoning on situations based on a set theoretical framework. This model has been instantiated using the rough sets theory. The proposed approaches and models have been implemented in prototypical systems. Their capabilities in improving SA in real applications have been evaluated with typical methodologies used for SA systems. [edited by Author]