Article Preview
Top1 Introduction
Wireless sensor networks (WSNs), characteristically comprising of huge number of collaborative and randomly deployed sensors nodes (SNs) which are widely used in various environmental monitoring applications (Akyildiz et al., 2002). The SNs have limited resource competencies in terms of sensing, transmission and computation and are usually functioned by batteries in a severe environment with non-replaceable power supply. Since most of the energy is consumed in sensing, communication and computation, which makes the network susceptible to failure. Several continuous monitoring applications like military surveillance, habitat monitoring, vehicle automation, industrial control systems, surveillance, office and home automation, traffic-monitoring water supplies, forest-fire monitoring, pollution monitoring, etc. awaits the SNs to be long lasting as they operate unattended. Consequently, the major issues and challenges in designing a sensor network are energy management and improving the network lifetime.
Various other sensor related factors that make the network more dynamic are location of SNs, their deployment and connectivity and most notably continuous drop in sensor network residual energy levels. Usually, the SNs collects data values over a time-period and send the aggregated data to the CHs or directly to BS periodically (Yick et al., 2008). To achieve precise data measurements, many data collection technique allows SNs to transmit all the data aggregates periodically. Due to excessive transmissions, such data collection techniques leads to high-energy drain that consequently results in early sensor failure.
In this context, data aggregation has emerge out to be an intelligent technique, where the data from various SNs are accumulated at intermediate SNs (or CHs), thereby reducing the number of packets to be sent to the BS. Moreover, in cluster based sensor networks, CHs consume more energy as they perform operations like data aggregating within cluster and then they transmit the aggregated data to the BS. Therefore, the appropriate selection of CHs in the sensor networks plays an essential role in prolonging the lifetime (Jain et al., 2019; Jain et al., 2020). Data aggregation in the extremely mutable data of time sequence and thus reduces the duplicate data transmissions resulting in huge energy preservation and better quality performance (Gupta & Sikka 2015; Jain & Bhola 2018).
Another crucial challenge for sensor network is dynamic aspects in reference to temporospatial variation detected in the course of sensed data because of recurrently varying environmental characteristics. Due to the random deployment of sensor network, the sensing regions may overlap in SNs and the gradual variation of the data is observed at a SN over time. The measurements either can be spatially correlated or temporally correlated, which results in the collection of abundant amount redundant information among the observed data.
Deploying an improved aggregation technique in sensor network will not only minimises redundant data forwarding, it also conserves abundant of energy from other sensor modules such as radio subsystem for communicating, sensing, and computing measure. Therefore, it is required to incorporate an effective data redundancy technique to avoid spatial correlation and temporal correlation over redundant data transmissions to ensure the durability of such continuous networking applications (Varun et al., 2004). In this work, we have presented a data aggregation technique for homogeneous application set-ups in sensor networks. For this, we have employed two ways of model generation for reducing Correlated Spatial-Temporal Data in cluster based sensor networks. One at the SNs and the other at the CHs. In this respect, we have proposed two types of data filtration, first in SNs for determining Temporal Redundancies (TRs) in data readings by both Relative Deviation (RD) and Adaptive Frame Method (AFM), and second in CHs for determining spatial redundancies (SRs) by both RD and AFM.
The remaining of this paper is systematized as below: In Section II, we present the review the related work in this direction. We represent the network architecture and describe the details of data filtration for determining temporal redundancies and spatial redundancies in Section III. Section IV presents a thorough simulations and experimental results followed by comparisons of our solution with the existing ones. Finally, we conclude this paper in Section V.