Spatio-Temporal Probabilistic Query Generation Models and Sink-Attributes Analysis in Energy Efficient Wireless Sensor Networks

Abstract

Rapid advancement in Micro-Electronic-Mechanical-Systems (MEMS) and distributed computing infrastructure along with compactness and economic viability in IC technology has accelerated the versatile growth and deployment of wireless sensors networks (WSNs). Owing to its sensing and subsequent parameters estimation abilities while maintaining higher spatial resolution led to a prominent position for WSNs in the networking paradigm. The main contribution of this thesis is to provide a modeling and analysis based on probabilistic framework for varieties of query generation scenarios in WSNs. In broad regime of query based WSNs, the query generation dynamics owe significantly to the associated spatial and/or temporal parameters. To encompass varying degree of uncertainties associated with spatio-temporal parameters; these parameters are treated as Fuzzified intervals, thus address quantum of uncertainties with the finest approximation or accuracy. Further, to ensure reliable network operation having fairly uniform coverage over the stipulated lifetime; usage of varieties of clustering schemes, sink attributes and spatial-fusion concept are explored. From operational aspects; the most important issues in wireless sensor networks (WSNs) are the coverage and the efficient usage of sensor nodes limited energy reserve. Irrespective of the applications served; owing to difficulty associated with battery replenishment, proper energy usage has been at centre-stage in WSNs operations, that ultimately influence the lifetime of WSNs. In this thesis; during various case-studies, square shape service-areas of varying area-dimension, different sensor node specifications and hierarchical network architecture are considered. These cases differ in terms of sink-attributes such as single/multiple, and stationary/portable. Sink is an important interface between the end user and the remote entities (sensor nodes), thus the proposed algorithms are formulated on considering sink at the center-stage. Further, the performance of WSNs also depends upon the topological structure of the network, usually it is hierarchical one. To realize hierarchical structure; the usage of clustering schemes namely static k-means (SKM), static fuzzy c-means (SFCM), dynamic k-means (DKM) and dynamic fuzzy cmeans (DFCM) are explored for the clusters formation and the subsequent selection of cluster heads (CHs). iiiSelection of CHs is done using residual energy status (RES) of participating sensor nodes. Cartesian-coordinate of these sensor nodes appropriately weighted with RES estimate decides the energy-centroid (EC) location for each cluster. The Euclidean distance measure between a sink and the ECs is used to identify new appropriate location for sink while complying with principal motive of energy conservation. Initially, during few query generation scenarios; different quadrants of the service-area observe distinct pattern of query spatial distribution. These query dissemination patterns are modelled using amplitude and angle modulated vectors. Later, the probabilistic approach is used to model the query generation process. The parameters of probabilistic models are regulated using the associated spatio-temporal aspects. In this thesis; usage of uniform probability mass function (PMF) and Poisson PMF models are investigated and analyzed to replicate query generation process. Further, the lifetime of WSNs depends upon sensor nodes energy dissipation pattern that is nonuniform in terms of spatial distribution over any short epochs. Thus, integrating the spatio-temporal aspects with the Poisson PMF model appears more reasonable. In mainstream probabilistic models; the associated control parameters are treated as crisp numbers, which fail to encompass uncertainties associated with the modelled parameters. To incorporate these uncertainties, Fuzzified interval-bound values of spatiotemporal parameters are considered to model the control parameter in Poisson PMF expression. Further, exploring the solutions in higher-dimensional space always entrust its superiority over the solutions that are derived from lower-dimensional space. This rationale is exploited using the concept of spatial (quadrants)-fusion in anticipation of improved profile of network performance measures. With these motivations, the thesis explores: (1) uses of energy efficient clustering schemes, (2) incorporation of spatio-temporal parameters uncertainties into probabilistic model of query generation using fuzzy-intervals bound, (3) importance of sink attributes, and (4) exploiting heuristic framework based on spatial-fusion concept to enhance network lifetime or to meet desired service norms over the stipulated network lifetime. For various network surveillance scenarios; the performance measures namely average residual energy status (ARES) of entire sensor network, critical residual energy status (CRES) of individual nodes as well as that of entire sensor network, fraction of sensor nodes attaining CRES mark and the network service-time-duration (STD) are estimated and analyzed.