Data Processing in the Emerging Internet of Things

Abstract

The emergence IoT is rapidly gaining ground in our modern society, aiming to improve the quality of life by connecting many smart devices, technologies and applications, for the purpose of exchanging data over the Internet. IoT devices will generate huge volumes of data in a rapid period of time and therefore require scalable solutions for dynamic and real-time processing of the generated data. Such solutions should provide a high level of accurate and reliable data for decision making. This requires data fusion, which is an efficient way for optimal use of a huge volume of data from multiple sources. We consider in this thesis the integration of IoT with edge, fog and cloud computing, the efficiency of data processing and fusion in terms of credibility, reliability, conflict, latency, and we propose several solutions. The first concerns the efficient processing of data in edge computing, which enables sophisticated services. The second approach is a hybrid computing-based IoT data management and control platform that enables heterogeneous resources, reliable connectivity and mobility, provides security, and contains services to merge data. heterogeneous. Numerical analysis and simulation results show that the proposed solutions allow significant savings in terms of energy consumption and reduction of lead times. The thesis also considers the state estimation in the average level of data fusion. We provide an improved distributed particulate filter algorithm to process target tracking in wireless sensor networks. It increases the estimation accuracy of the particulate filter, improves the efficiency of particle sampling, and improves the estimation performance. The simulation and numerical analysis results show the superiority of the proposed approach in terms of root mean square error and scalability. We have studied the problem of data fusion at the decision-making level. Reliability and conflicts are taken into account in our method by considering the information lifetime, the distance between sensors and features, reducing the computation and using combination rules based on the base probability assignment . This makes it possible to represent uncertain information or to quantify the similarity between two bodies of evidence. We compared the proposed solution with state-of-the-art data fusion methods, and using both benchmark data simulation and an actual data set from an intelligent building testbed. The results show that our solution outperforms methods in terms of reliability, accuracy and conflict handling.