Data aggregation for wireless sensor networks: an energy-efficient approach

Abstract

As a major component of Internet of Things (IoT), pervasive wireless sensor networks (WSNs) are becoming ubiquitous in all aspects of life such as environmental monitoring, military, intelligent transportation. Data gathering is a crucial yet energy-consuming task for WSNs. More and more sensing devices are deployed, which results in a huge amount of data to be transmitted. Usually, the sensor node is equipped with tiny, non-rechargeable battery and limited computation capability. Consequently, it’s imperative to improve energy efficiency and prolong network lifetime for WSNs powered by non-rechargeable batteries. In addition to non-rechargeable battery, energy harvesting (EH) has emerged to be a promising power supply to WSNs, which may unlock the possibility of perpetual operation for WSNs. In practical environment, data redundancy and data correlation commonly exist, which motivates the development of data aggregation techniques for both battery powered and energy harvesting WSNs (EH-WSNs). Data aggregation can be categorized as function-based and correlation-based. This thesis aims to develop both function-based and correlation-based data aggregation algorithms for conventional WSNs powered by non-rechargeable battery to achieve energy efficiency and prolong network lifetime. Further, this thesis applies data aggregation into EH-WSNs to achieve perpetual operation and satisfying application requirements. First of all, this thesis proposes a function-based data aggregation algorithm for a single-hop sensor network. In certain applications such as alarm systems, only the maximum value is required. Therefore, function-based data aggregation can be used to reduce energy consumption as well as transmission delay. A priority-aware hybrid scheduling scheme is proposed to achieve fast decentralized energy-efficient MAX function computation in a single-hop network. In our proposed scheme, priority-aware TDMA scheme is used to schedule the transmission. CDMA is used to recover information whenever a collision occurs. The proposed scheme helps to reduce the number of transmissions and delay for MAX computation with good scalibility. Secondly, this thesis focuses on correlation-based data aggregation algorithms based on compressive sensing (CS). Leveraging the load balancing features of CS, a novel hybrid CS based data gathering algorithm with a dense measurement matrix for tree-based multi-hop data gathering network is developed. CS technique can guarantee the full recovery of sensor data while imposing simple computation load on the sensor. An optimization problem is formulated and solved by approximation algorithms. The proposed scheme significantly reduces total energy consumption and data gathering latency and dramatically improves network lifetime compared with benchmark algorithms. The sensor network may experience changing environment where deterministic routing path is vulnerable to attacks and is less energy-efficient. In this case, it would be beneficial to allow a certain amount of exploration in the routing path. A cost-aware stochastic compressive data gathering algorithm is developed for multi-hop WSNs. The mathematical Markov chain based model is used to analyze the stochastic compressive data gathering process. A sparse random measurement matrix with optimized compression probability is constructed to minimize expected total cost while guaranteeing the recovery accuracy. Extensive simulations on both synthetic and real data show that the proposed algorithm requires less total expected cost to achieve a certain level of accuracy compared with benchmark algorithm. Finally, the proposed algorithm prolongs network lifetime for a given recovery accuracy. Finally, this thesis proposes a harvesting-aware compressed data gathering algorithm, in which the heterogeneous EH rates of the nodes are taken into consideration in order to maximize information throughput while maintaining sustainable operation. An optimization problem is formulated and solved by approximation algorithms. Our proposed algorithm can significantly improve information throughput while maintaining sustainable operation.