Desktop version

Home arrow Computer Science

  • Increase font
  • Decrease font

<<   CONTENTS   >>


In this section, the working principle of the FC-DR protocol has been presented. The basic architecture of clustering-based IoT-WSN model is depicted in Figure 4.2. The presented FC-DR method operates on three major stages: node clustering, data collection, and data aggregation. Initially, the IoT sensor nodes perform FC process to select CHs and organize clusters. In the second level, the CMs observe the environment and forward the data to the CH. Finally, in the third level, the CHs perform data aggregation using EBLC technique.

Fuzzy-Based Clustering Process

For reduced energy utilization, the cluster formation process plays an important part. It applies k-means clustering technique to the cluster formed. The counts of datasets are

Architecture of IoT-WSN

FIGURE 4.2 Architecture of IoT-WSN.

separated into /с-clusters that utilize these techniques. A value of к is estimated as given in equation (4.1):

where n is the node count of sensor nodes, D is the network size, and is the average distance of every node for the BS. Utilizing the Euclidean distance, the distance among all of the sensors nodes for the entire clusters center is planned as given in equation (4.2):

where Xn2CC indicates node’s distance from cluster center, Xj signifies the direct node j, and Xcc is the cluster center.

The CH selection takes place using remaining energy, communication rate between a node and its neighboring node, link quality, restart value, and count of neighboring nodes, and node marginality, respectively. Remaining Energy Level

Energy is a significant resource in WSN. The CHs are the nodes that utilize and further energy than CMs if they contain aggregating, computing, and routing information. The remaining energy is calculated as given in equation (4.3):

where E0 and Ec are the primary energy and the energy utilized with the node, respectively, and E,. is the remaining energy of a standard node. Communication Rate

The broadcasting message utilizes the energy that is simultaneously a square of the distance between the applicant and source nodes. A charge of communication rate is determined as given in equation (4.4):

where dmg signifies the average distance among the neighboring nodes and d0 is the transmitting range of the nodes. Link Quality

In WSN, the disappearing channel is usually arbitrary and time-different. Until a receiver does not examine the signal properly, a rebroadcast can happen and it needs further energy dissipation of the broadcaster. Thus, a link quality should be estimated for accomplishing energy competence. A link quality can be computed as given in equation (4.5):

where Qmax and Qmin are the maximal and minimal count of rebroadcast from the neighborhood, respectively, and Q, indicates the entire rebroadcast number among the neighbors and the node. Restart Value

A node is essentially an embedding method. Occasionally, the methods affect a software or hardware fault. For solving these issues, the watchdog circuit is employed for restarting the PC scheme to ensure the node continues functioning. So, the frequent restart is utilizing any further energy. A value of restart value is estimated utilizing equation (4.6):

where Smax and Smj are the maximal and minimal restart values obtained from neighboring, nodes, respectively, and S0 indicates the entire restart value behind the WSN has been arranged. Node Degree

A principle is that the nearer the neighboring nodes, the more effective the node and the greater possibility of becoming a CH. The node degree can be calculated as given in equation (4.7):

where D, indicates the count of neighboring nodes and D0 is the better count of neighboring nodes. Node Marginality

A division of coverage area will incorporate absence of nodes when a node is placed at the boundary of the observing region. A node simply encloses a limited area. Thus, the entire count of CHs in the network gets enhanced. Node marginality is determined as follows:

where q indicates the quadrant number.

In fuzzy surroundings, fuzzy analytic hierarchy process (AHP) is a valuable method under several conditions of decision-making. According to the function objectives and user preferences, the weights are assigned for all conditions in AHP. A fuzzy AHP technique depends upon the following stages:

1. Creation of pairwise comparative decision matrix

A pairwise comparative matrices are provided as follows:

where y„ = 1, y? = 11 у

2. Normalization of decision matrix as computed in equation (4.10)

3. Weighted normalized decision matrix as provided in equation (4.11)

where n is the criterion number.

Data Aggregation Process

The EBLC method such as SZ performs high-performance computing (HPC) functions. These compression methods have been presented to manage the massive counts of data created in the implementation of HPC functions. The actual SZ reduces input data records, which are in binary formats and contain several data shapes as well as types. It is presented for adapting the SZ technique to IoT tools by considering the floating point data type and removing other types that create the code tiny in size and are simple for compiling on small tools. Besides, the technique was modified for taking a ID arrangement of float sensor data as input and replacing a byte range, which exists to be broadcasted for the edge node. After selecting, SZ for IoT functions as follows:

  • • SZ access the encoding of multivariate time series, including different features through several scales.
  • • SZ permits managing the data loss by utilizing an error bound.
  • • SZ takes superior compression ratio compared to the multidimensional change field compression methods

It is regarded that the information are broadcasted for the edge behind all periods P of time f. The gathered information is in the structure of M X N array, where M indicates the count of readings and N refers to the count of features. In the beginning, the 2D array is changed to the ID array. Thereafter, the flattened array compresses utilizing the lossy SZ method. Finally, the outcome binary array is broadcasted for the edge. It is noticeable that the adaption has been completed through removing the essential performances from the actual SZ for making it suitable on wearable and resource-limited tools.

An SZ compressing technique begins with reducing the ID array utilizing adaptive curve-fitting methods. The optimal fit stage uses three forecast methods: preceding neighbor fitting (PNF), linear curve-fitting (LCF), and quadratic curve-fitting (QCF). A variation among the three methods is found in the count of precursor data points needed to suit the actual rate. An adapted method is the one that gives the nearer estimate. It is to be noticed that the suitable information is changed into integer quantization factors and encoding that utilize Huffman tree. If no forecast methods in the curve-fitting stage ensure the error limit, the data point is clear as uncertain and the next encoding examines the IEEE 754 binary illustration. In the error bound, an absolute error bound (AEB) is utilized, implying that the compressing or decompressing faults are restricted to be in an AEB. For example, when the rate of data point is regarded to be X, an AEB of 10"1 implies, and thus the decompressed rate might be in the series [X-10"1, X + 10-1].

<<   CONTENTS   >>

Related topics