Managing big lot data in smart environments

Abstract

Data quality is crucial in IoT-based smart environments, where the reliability and ac- curacy of information collected from interconnected devices significantly influence system effectiveness and efficiency. Multisensor data fusion technique has emerged as a powerful tool for managing imperfect data from heterogeneous sources, thereby enhancing opera- tional efficiency and enabling effective decision-making. The Dempster-Shafer (DS) theory of evidence provides a robust and flexible mathematical framework for modeling and fus- ing uncertain, imprecise, and incomplete data. However, Dempster's combination rule can lead to counterintuitive results when dealing with highly conflicting data sources. This thesis focuses on enhancing data quality within IoT-based smart environments by investigating data fusion techniques for managing heterogeneous IoT data. It specif- ically addresses the complexities arising from highly conflicting data sources within the Dempster-Shafer theory framework. Novel solutions are proposed in this work to over- come limitations associated with Dempster's combination rule and to improve the quality, reliability, and utility of data from multiple sensors. These solutions involve preprocessing the original evidence model by assigning weighting factors to evaluate the reliability of each information source, considering both uncertainty and conflict using various metrics. To demonstrate the validity and effectiveness of the proposed approaches, simulations are conducted across various domains, including fault diagnosis, IoT decision-making, and situational awareness within UAV systems. Additionally, a comparative analysis with several similar methods from existing literature is carried out to validate the efficiency and superiority of the proposed solutions in terms of conflict management effectiveness, convergence, fusion result reliability and decision accuracy. These findings contribute to achieving more robust and trustworthy outcomes in dealing with complex and conflicting data.