Wireless sensor networks (WSN) have played a role in various fields. In-network data processing is one of the most important and challenging techniques as it affects the key features of WSNs, which are energy consumption, nodes life circles and network performance. In the form of in-network processing, an intermediate node or aggregator will fuse or aggregate sensor data, which are collected from a group of sensors before transferring to the base station. The advantage of this approach is to minimize the amount of information transferred due to lack of computational resources.
This thesis introduces the development of a hybrid in-network data processing for WSNs to fulfil the WSNs constraints. An architecture for in-network data processing were proposed in clustering level, data compression level and data mining level. The Neighbour-aware Multipath Cluster Aggregation (NMCA) is designed in the clustering level, which combines cluster-based and multipath approaches to process different packet loss rates. The data compression schemes and Optimal Dynamic Huffman (ODH) algorithm compressed data in the cluster head for the compressed level. A semantic data mining for fire detection was designed for extracting information from the raw data by the semantic data-mining model is developed to improve data accuracy and extract the fire event in the simulation. A demo in-door location system with in-network data processing approach is built to test the performance of the energy reduction of our designed strategy. In conclusion, the added benefits that the technical work can provide for in-network data processing is discussed and specific contributions and future work are highlighted.
A Doctoral Thesis. Submitted in partial fulfilment of the requirements for the award of Doctor of Philosophy of Loughborough University.