Aggregation and Recovery Methods for Heterogeneous Data In IoT Applications

  • Evangelos Zimos ((PhD) Student)

Scriptie/masterproef: Doctoral Thesis


Recent advances on sensing hardware, wireless communications, cloud computing and data analysis have powered intelligent data acquisition systems, equipped with smart wireless sensor devices. Wireless sensor networks have thus brought new perspectives in our lives as they support the development of various applications related to smart cities, ambient-condition monitoring, air-pollution monitoring, smart energy metering and many more. These wireless devices gather data expressed by different marginal statistics— namely, data of heterogeneous types—and aggregate it to a central node. Designing data acquisition schemes for such systems is a challenging task. Firstly, the nodes need to optimize their energy resources, such that their battery depletion is avoided and the power autonomy of the system is achieved. Secondly, the energy resources of the entire network should be wisely managed such that all devices send a similar amount of information and therefore need similar energy resources. Thirdly, the sensed information should be recovered robustly at the sink without being significantly corrupted by imperfections of the wireless channel.
To deal with the energy constraints on the nodes, we propose code designs that significantly reduce the required transmission rate. Unlike existing schemes, our designs achieve superior in-network compression performance on the collected data by leveraging the statistical dependencies among heterogeneous sensor data during recovery. To efficiently capture different types of data dependencies, we propose the use of copula functions, a flexible statistical model that provides accurate multivariate expressions based on different marginal statistics. For small-to-medium scale networks, we introduce a code design that combines multiterminal source coding with copula regression and predictive coding mechanisms. For large-scale setups, we propose three new compressive data gathering schemes that not only provide low transmission rate, but also result in balanced energy consumption in the system.
Datum Prijs26 sep 2017
Toekennende instantie
  • Vrije Universiteit Brussel
BegeleiderNikolaos Deligiannis (Promotor), Adrian Munteanu (Promotor), Ann Nowe (Jury), Rik Pintelon (Jury), Martin Timmerman (Jury), Jan Lemeire (Jury), Miguel Rodrigues (Jury) & Eli De Poorter (Jury)

Citeer dit