ENERGY EFFICIENT COMPRESSIVE DATA GATHERING INWIRELESS SENSOR NETWORKS

Abstract

Wireless sensor networks (WSN) are most preferable and widely used application driven technology that helps to remotely monitor or control the difficult task. This has made life simple to a great extent. Wireless sensor network is one of the technology that has boosted the industrial automation to large extent. The advancement in wireless sensor network (WSN) is rooted in modern fabrication technology which helps to fabricate small match box-sized sensor node (SN)s with the ability to sense the physical environment, pre-process the data locally and communicate the information to a distant sink with limited available battery power. In addition to these basic requirements, some additional features such as global positioning system (GPS) and many more as per requirement of the application, may also be embedded into match box-sized sensor node. Although, it has paved path for automation, but due to limited processing power and stringent energy constraints, the execution of communication protocols which consumes major portion of energy in a sensor node has become complicated. Earlier communication systems involved transmitters which did not have any limitations of either power or computation, whereas receivers were limited powered devices. Therefore, communication protocols designed were complex in encoding and easy to decode. But, the scenario is just reverse in wireless sensor networks here the sensor nodes which are transmitters have limited energy and computational power, whereas the sink or the receiver is not resource constrained. This has paved path for the development of efficient communication protocols constrained on limited computational and power resources and has attracted the attention of communication and network engineers these days. In this thesis, we investigated energy preservation in wireless sensor network. Wireless sensor networks are usually deployed in unaccessible terrains either for event detection application or for continuous monitoring application. In event detection applications, the sensed signal is a sparse signal whereas in continuous monitoring applications, the sensors sample the physical environment at very high rate such that the rate of variation of the environment is comparatively low. Thus, temporally correlated redundant data is generated in wireless sensor network. Apart from this, due to low cost and prone to faliure nature of sensor node, these are deployed redundantly in the area under observation. Due to redundant deployment, spatial redundancy exists in wireless sensor network. This redundant data cause signal to berepresented sparsely in some orthonormal domain. Due to sparse nature of signal we propose to exploit data redundancy and compress the effective data to be transmitted from nodes to the sink. The main task of WSN is data gathering and data aggregation. This is an active area of research that has motivated this thesis. In this thesis, we propose to exploit joint sparsity of the signal to compress the data. For this, we proposed 2-D distributed compressive sensing (DCS) which compresses the data and reduces the in-network traffic to a large extent, thereby preserving energy of the wireless sensor network which in turn increases the network lifetime. In addition to energy saving, compressive data gathering (CDG) using DCS helps to balance the load among sensor nodes evenly in multi-hop communication. Thus, load balancing helps to avoid hot spot or the energy hole problem caused due to failure of bottle neck nodes. Simulations are performed using real sensor readings of different datasets. Results indicate that by exploiting both temporal and spatial correlation simultaneously large amount of energy is preserved as compared to traditional compression techniques which have huge computational complexity. The proposed distributed compressive sensing (DCS) involves compression of the data with minimum computational complexity. As in DCS, the data compression involves only matrix multiplication operations whereas decoding of the original signal at the sink or the receiver involves reconstruction from an under-determined set of linear equation which is an ill posed mathematical problem that results infinitely many solutions. Obtaining an unique solution to such system of equation requires additional constraint to be satisfied. This makes reconstruction a non-polynomial-times hard (NP-Hard) problem. The performance of the proposed 2-D DCS measured in terms of average mean square error (MSE). Using existing various reconstruction algorithms. The energy efficient data gathering is attained at the cost of greater reconstruction delay. Further, for compressing the data using the proposed DCS we used different measurement matrices. The performances of various measurement matrices are simulated with real sensor readings and are measured in terms of percentage of exact recovery and MSE. Also, the delay in reconstruction using different measurement matrix is found. Results reveal that random discrete cosine transform (DCT) matrix gives better reconstruction with negligible MSE and lesser reconstruction delay. Lastly, to preserve energy in wireless sensor network, we propose an energy efficient possibilistic clustering with load balanced routing which increases network lifetime (NL) of the WSN as compared to possibilistic c-means (PCM) and helps to generate better clusters. The proposed energy efficient possibilistic clustering with load balanced routing is an hierarchical routingmechanism. In the first phase of clustering, the cluster head (CH)s or the gateway node locations are selected based on the possibilistic clustering phenomenon and then in the next phase each sensor node is assigned membership to any particular cluster so as to minimize the energy consumption in transmitting data from member sensor node to the cluster head via multi-hop communication. The energy consumption model is based on direct line of sightcommunication in which the energy consumed in transmitting is proportional to the square of the distance. The routing of data from sensor node to cluster head involves in-network data compression using our proposed DCS thereby enabling load balanced routing of data. Simulation results reveal improvement in network lifetime on account of clustering. Further, network lifetime of the proposed clustering mechanism is better than possibilistic clustering measured in terms of first node dies (FND) and half of the nodes alive (HNA).