Research On Compressive Sensing Based Data Gathering In Wireless Sensor Networks

Wireless sensor networks (WSNs) is one of the main constitution of Internet of Things (IoT). WSNs have been widely used in environmental monitoring and protec-tion, medical care, military field, target tracking and other related areas. However, WS-N is a resource-constrained and error-prone network, extremely large amount of data transmission hinders the applicability and the reliability of the large-scale WSNs de-ployment. Leveraging the spatial-temporal properties in sensory data from real deploy-ments, in-network compression is an essential technique to reduce the amount of data transmission preserving relatively high reconstruction accuracy in the sink. Traditional data compression techniques need large amount of data exchanges among sensor nodes and/or high computational complexity, which is not suitable to recourse-constrained sensor networks. In recent years, Compressed Sensing (CS) theory has been proposed, which has opened up a new research avenue for the data sampling and information ac-quisition. Compared with traditional data sampling and compression method, it can accurately recover the sensory data and has very simple encoder.In recent years, many CS based data gathering schemes have been proposed, which has opened up a new research avenue for in-network compression. However, existing schemes naively applied CS to gathering compressive sensory data in WSNs. CS theory is also in constant development, to truly achieve sampling and compression at the same time, improve the network performance and reduce network bottlenecks, etc. There are many challenges need to be studied.In this dissertation, we carried out the related researches on CS based compressive data gathering in WSNs. Our main contributions are follows:For most existing methods only random projection spatial sensory data, it leads to poor quality of data recovery and relative highly transmission cost. In this disserta-tion, we propose a distributed spatial-temporal compressive data gathering in large-scale WSNs. In our scheme, we first present a distributed spatial-temporal sensory data parti-tion model to improve the compressibility of gathering data. Sparse and dense random projections are used for compressing and gathering different components of our sensory data partition model to obtain the same projection process. We also exploit cluster-based routing strategy to gather CS measurement and reduce energy consumption.For existing CS based data gathering methods adopts tree type routing scheme, it requires a large number of sensors to participate each CS measurement gathering which leads to waste a lot of energy. In this dissertation, we present an energy efficient cluster-ing routing data gathering scheme for large-scale wireless sensor networks. To obtain the optimal number of clusters, we first formulate an energy consumption model to ob-tain the optimal number of clusters. Second, we design an efficient non-probabilistic dynamic clustering (ECDC) scheme to make each clustering process have the same number of clusters and guarantee all cluster heads roughly uniformly distributed. Our scheme can make the network energy consumption very uniformly and significantly prolongs the sensor network’s lifetime compared with tree routing strategy.Consider the sparse level of measurement matrix affect the transmission cost dur-ing each CS measurement gathering. In this dissertation, we propose a sparsest random scheduling for compressive data gathering scheme, which decreases each measurement transmission cost from O(N) to O(log(N)) without increasing the number of CS mea-surements as well. In our scheme, we present a sparsest measurement matrix where each row has only one nonzero entry. To satisfy the restricted isometric property (RIP), we propose a design method for representation basis, which is properly generated ac-cording to the sparsest measurement matrix and sensory data.In this dissertation, we carry out our works from three aspects:Establishing spatial-temporal sensory data model to improve the compressibility of sensory data to reduce the number of measurements; Establishing a dynamic clustering routing data gathering scheme to reduce transmission cost; Designing the representation basis for sensory data to achieve the sparsest random scheduling data gathering. Experimental results show that our propose methods can effectively reduce the number of packets in the network transmission and effectively recover sensory data.