nodes (1)

<sup>A</sup> (2)

to have a higher value for this parameter to become a permanent cluster head.

NodeCentrality ¼

compute its probability to become to a permanent cluster head.

earlier and joins. Thus overlapping of clustering is avoided.

is a member of the set 2-hop-nbr. The variable "A" represents area of the network.

formed on 2-hop coverage and hence the aggregation can be done on a larger scale compared to the conventional 1-hop clustering approaches.

The cluster heads have to report the aggregated data to the sink. Unlike LEACH and many conventional algorithms, the proposed algorithm inherits the presence of multi-hop relay between the cluster head and the sink. This multi-hop relay favors the selection of any one path based on certain probability among multiple choices and the selected path is not necessarily to be an optimal path as followed in many approaches. The repetitive employment of a few popular paths which have been identified as efficient paths causes energy depletion of nodes on these paths and pushes them die soon. Our idea of selecting less used or unused


Table 1. Fuzzy sets for input and output variables.

paths can effectively contribute to the distribution of energy consumption and prolong the lifetime of sensor nodes.

#### 2.3. Performance evaluation of EAFCA

The performance of the algorithm is evaluated under three different scenarios. In scenario 1, the sink is positioned at the center and 100 sensor nodes are deployed. In scenario 2, the sink is positioned at the center while the number of sensor nodes deployed was doubled as 200 to test the scalability of the network. In scenario 3, the sink is located outside the predefined WSN boundaries and the network size is maintained as in scenario 2. The performance of EAFCA algorithm is compared against the benchmarking protocols namely, low energy adaptive clustering hierarchy (LEACH), energy aware clustering protocol using fuzzy logic (ECPF) and energy aware clustering scheme with transmission power control for sensor networks (EACLE).

The metrics employed for computing the lifetime of sensor networks are, first node dies (FND) and half of the nodes alive (HNA) metrics. These two metrics are widely adopted based on the viewpoint whether the energy depletion of the very first node in the network or half of the nodes indicate the death of the network. From the simulation results, it has been observed that the proposed algorithm shows 88% energy improvement compared to LEACH, 46% of improvement with respect to EACLE and 30% of improvement with respect to ECPF.

It is to be observed that LEACH shows the poorest performance among the selected clustering approaches since it does not re-elect cluster heads from the energy perspective and it continues to be a probabilistic model. EACLE shows some improvement in energy consumption as studied from the simulation results. The gain in energy efficiency is achieved in EACLE since it employs multiple paths for inter-cluster traffic and postpones the death of sensor nodes. ECPF claims more energy improvement since it adopts a fuzzy based cluster head election. The results observed across both FND and HNA metric confirms this claim.

The proposed EAFCA reduces energy efficiency by considering necessary and sufficient parameters for a cluster head election and assumes a feasible configuration in which 2-hop coverage is given for every cluster head and multi-hop relay is done for inter-cluster communication. The results demonstrate that EAFCA keeps WSN functioning for longer time than the other approaches.

This work stands as a representative for cognitive and effective cluster head election process. Such strategies expose the sensor networks and its applications to the emerging era of explorations and can be eventually commercialized.

## 3. Efficient energy harvesting assisted clustering (EEHC) algorithm

## 3.1. Energy harvesting in wireless sensor networks

paths can effectively contribute to the distribution of energy consumption and prolong the

Energy 2-hop ND Node centrality Chance Energy Low Low Far Very weak Low Low Low Medium Weak Low Low Low Close Little weak Low Low Medium Far Weak Low Low Medium Medium Little weak Low Low Medium Close Little weak Low Low High Far Little weak Low Low High Medium Little weak Low Low High Close Medium Low Medium Low Far Little weak Medium Medium Low Medium Little medium Medium Medium Low Close Medium Medium Medium Medium Far Little medium Medium Medium Medium Medium Medium Medium Medium Medium Close High medium Medium Medium High Far Medium Medium Medium High Medium High medium Medium Medium High Close Little strong Medium High Low Far Medium High High Low Medium High medium High High Low Close Little strong High High Medium Far High medium High High Medium Medium Little strong High High Medium Close Strong High High High Far Little strong High High High Medium Strong High High High Close Very strong High

The performance of the algorithm is evaluated under three different scenarios. In scenario 1, the sink is positioned at the center and 100 sensor nodes are deployed. In scenario 2, the sink is positioned at the center while the number of sensor nodes deployed was doubled as 200 to test

lifetime of sensor nodes.

2.3. Performance evaluation of EAFCA

Table 1. Fuzzy sets for input and output variables.

148 Wireless Sensor Networks - Insights and Innovations

Lifetime of wireless sensor applications depends upon the lifetime of the sensor nodes which are constrained by their energy resources. This can be managed by the use of energy harvesting, utilizing ambient sources to prolong the life of the batteries in wireless sensor nodes. The efficiency of this approach depends upon how much energy is harvested. This can majorly influence the lifetime of sensor nodes and in turn that of the sensor network. In our efficient energy harvesting assisted clustering (EEHC) for wireless sensor networks, the effective energy harvesting for wireless sensor networks is experimented and studied through an efficient energy budgeting. The measurement of energy consumption, energy budgeting and energy harvesting are presented as follows.

#### 3.2. Energy consumption

A sensor node consumes energy during sensing the data and forwarding it to the cluster head. A cluster head consumes the energy during the reception, data aggregation and forwarding the aggregated data. The energy consumption of a cluster member to sense and transmit 1-bit of information to the cluster head is estimated in Eq. (3):

$$\mathbf{E\_{SN}} = \mathbf{S\_{ensing}} + \mathbf{E\_{tx}} \tag{3}$$

Assuming a sensing rate of "x," the total data sensed and transmitted by "n" cluster members in a time period "t" is estimated as given in Eq. (4).

$$\mathbf{E\_{CM}} = (\mathbf{n.x.t}).\mathbf{E\_{SN}}\tag{4}$$

Since the maximum number of cluster members are located at 1-hop distance to the cluster head, it is assumed that the data sensed at time "t" is transmitted to the cluster head within the same interval. Suppose a cluster head collects L-bit length of data at time "t," (i.e., L= x.t.n) then the total energy conservation for data reception, aggregation and forwarding in that CH across time period "t" is estimated as given in Eq. (5).

$$E\_{CH} = n.x.t.E\_{rx} + n.x.t.E\_{DA} + \frac{n.x.t}{\alpha}.E\_{trr} \tag{5}$$

where α stands for aggregation ratio.

The total energy consumed in time "t" for a cluster is given in Eq. (6).

$$\mathbf{E\_{c}} = \mathbf{E\_{CH}} + \mathbf{E\_{CM}} \tag{6}$$

For a time slot "t," the entire cluster, i.e., all the cluster nodes including the cluster head should harvest the energy equal to that of the estimated energy. Suppose there are "n" cluster members and a cluster head, then the energy that is required to be harvested by a sensor node in a cluster is given by the Eq. (7).

$$E\_h = \frac{E\_c}{n+1} \tag{7}$$

#### 3.3. Energy harvestings

For every time interval "T" between time "t1" and "t1+T," the harvested energy is calculated as given in Eq. (8).

$$E(t\_1 \text{ to } t\_1 + T) = E\_t(n) \, + \tau \int\_{t\_1}^{t\_{1+T}} E\_h(n, t) \, dt - \int\_{t\_1}^{t\_{1+T}} E\_l(n, t) \, dt \tag{8}$$

In Eq. (8), the three components represent the energy of the node at starting time "t1," energy harvested at time interval "T" and energy leakage during this interval. The factor "τ" represents charging efficiency. All the sensor nodes are provided with the storage buffers to store the harvested energy.

## 3.4. Energy budget

3.2. Energy consumption

150 Wireless Sensor Networks - Insights and Innovations

of information to the cluster head is estimated in Eq. (3):

in a time period "t" is estimated as given in Eq. (4).

across time period "t" is estimated as given in Eq. (5).

where α stands for aggregation ratio.

cluster is given by the Eq. (7).

3.3. Energy harvestings

given in Eq. (8).

the harvested energy.

A sensor node consumes energy during sensing the data and forwarding it to the cluster head. A cluster head consumes the energy during the reception, data aggregation and forwarding the aggregated data. The energy consumption of a cluster member to sense and transmit 1-bit

Assuming a sensing rate of "x," the total data sensed and transmitted by "n" cluster members

Since the maximum number of cluster members are located at 1-hop distance to the cluster head, it is assumed that the data sensed at time "t" is transmitted to the cluster head within the same interval. Suppose a cluster head collects L-bit length of data at time "t," (i.e., L= x.t.n) then the total energy conservation for data reception, aggregation and forwarding in that CH

For a time slot "t," the entire cluster, i.e., all the cluster nodes including the cluster head should harvest the energy equal to that of the estimated energy. Suppose there are "n" cluster members and a cluster head, then the energy that is required to be harvested by a sensor node in a

Eh <sup>¼</sup> Ec

For every time interval "T" between time "t1" and "t1+T," the harvested energy is calculated as

ð<sup>t</sup>1þ<sup>T</sup> t1

In Eq. (8), the three components represent the energy of the node at starting time "t1," energy harvested at time interval "T" and energy leakage during this interval. The factor "τ" represents charging efficiency. All the sensor nodes are provided with the storage buffers to store

Ehð Þ n; t dt �

ð<sup>t</sup>1þ<sup>T</sup> t1

ECH ¼ n:x:t:Erx þ n:x:t:EDA þ

The total energy consumed in time "t" for a cluster is given in Eq. (6).

E tð Þ¼ <sup>1</sup> to t<sup>1</sup> þ T EtðÞþ n τ

ESN ¼ Sensing þ Etx (3)

ECM ¼ ð Þ n:x:t :ESN (4)

Ec ¼ ECH þ ECM (6)

<sup>n</sup> <sup>þ</sup> <sup>1</sup> (7)

Elð Þ n; t dt (8)

<sup>α</sup> :Etxr (5)

n:x:t

The energy consumed must be compensated by the energy harvested within the boundary of a cluster in any given time slot, i.e., the energy budget should harvest more energy than that of the energy consumed in every cluster periodically. An efficient energy budget should ensure that the energy consumption should not be increased than the energy harvested across any time slice.

### 3.5. Performance evaluation of EEHC

The performance of our EEHC has been compared with a modern clustering protocol named energy harvesting and information transmission protocol (EHITP) and the classical clustering protocol LEACH under three scenarios. In scenario 1, 100 sensor nodes are deployed in a region of 200 200 m<sup>2</sup> . In scenarios 2 and 3, the population of sensor nodes and area are doubled successively. The experimental results indicate that EEHC exhibits a mean improvement of 91 and 67% when compared to LEACH and EHITP, respectively.

Thus the harvesting can be made efficient through appropriate budgeting in wireless sensor networks and this budgeting further is influenced by the nature of the sensor applications and critical dynamic constraints.

## 4. Energy-efficient recursive clustering (EERC) algorithm

### 4.1. Event based clustering in wireless sensor networks

Generally, wireless sensor networks are employed for two purposes: continuous data monitoring and event monitoring. An example for the former category is a weather monitoring sensor network that measures temperature, moisture, etc. A typical event-based sensor network is habitat monitoring such as surveillance of wild animals and smart home applications. Except a few applications in the second category, most of the senor nodes are put on sleep mode in order to save power and the effective sleep/wakeup scheduling algorithms are required to determine the set of sensor nodes that can be scheduled to sleep with respect to time.

The energy-efficient recursive clustering (EERC) algorithm is an event-driven clustering algorithm, i.e., on the occurrence of an event, the clusters are formed to reduce energy dissemination, in a recursive fashion. The recursive clustering approach employs two stages of clustering process. The first stage of clustering is followed by further partitioning of clusters. Then CHs are elected from energy perspective.

#### 4.2. Recursive clustering approach

The recursive clustering approach employs two stages of clustering process. After the deployment of sensor nodes, the distance between the nodes is computed using Euclidean distance. Based on the distance, "k" number of clusters are formed which results in the first stage of clusters. Since the clustering process is modeled as recursive process further the "k" number of clusters is divided in "j" number of clusters based on the distance and interval between the nodes which leads to the second stage of clusters. A typical process of this two-stage clustering is pictorially represented as shown in Figure 3. After recursive clustering process, CH is elected based on energy levels and employing round robin scheduling algorithm. Based on the computations, the node with minimum turnaround time and high energy is elected as the CH among the nodes in the cluster.

Each node senses the data for every two rounds. After the completion of two rounds, turnaround time is calculated. The node having the minimum turnaround time is elected as the cluster head among the nodes in the cluster. The sensed data from each node is sent to the cluster head. In the cluster head, data is aggregated and sent to the base station by the multihop routing. The aggregated data in a cluster head leads to less transmission data, decrease in overheads and decrease in energy consumption.

## 4.3. Performance evaluation of EERC

The performance of our EERC has been evaluated against the conventional LEACH clustering approach under three scenarios through simulation. In scenario 1, the number of sensor node deployed is 100 and in order to test the scalability the scenario 2 and scenario 3 were considered. In scenario 2, 250 sensor nodes and in scenario 3, 500 sensor nodes have been considered.

From the results obtained for scenario 1, our EERC approach shows performance improvement when compared to the classic LEACH protocol. For 100 nodes, there is 23.85% increase

Figure 3. Recursive clustering in wireless sensor networks.

in lifetime, 0.287% decrease in energy consumption, 2.522% decrease in delay, 1.497% increase in transmission time, 1.8% increase in goodput and 0.56% decrease in overhead. In scenario 2, EERC shows performance improvement of 12.58% increase in lifetime, 0.402% decrease in energy consumption, 9.815% decrease in delay, 9.289% increase in transmission time, 0.524% increase in goodput and 0.554% decrease in overhead. In scenario 3, EERC exhibits performance improvement of 11.03% increase in lifetime, 0.619% decrease in energy consumption, 5.735% decrease in delay, 9.289% increase in transmission time, 0.524% increase in goodput and 0.554% decrease in overhead throughput.

Thus the recursive clustering technique gives considerable improvement across various performance factors, sustaining the equilibrium of the entire network.

## 5. Adaptive distributed clustering algorithm (ADCA)

## 5.1. Similarity measure in sensor data

is pictorially represented as shown in Figure 3. After recursive clustering process, CH is elected based on energy levels and employing round robin scheduling algorithm. Based on the computations, the node with minimum turnaround time and high energy is elected as the

Each node senses the data for every two rounds. After the completion of two rounds, turnaround time is calculated. The node having the minimum turnaround time is elected as the cluster head among the nodes in the cluster. The sensed data from each node is sent to the cluster head. In the cluster head, data is aggregated and sent to the base station by the multihop routing. The aggregated data in a cluster head leads to less transmission data, decrease in

The performance of our EERC has been evaluated against the conventional LEACH clustering approach under three scenarios through simulation. In scenario 1, the number of sensor node deployed is 100 and in order to test the scalability the scenario 2 and scenario 3 were considered. In scenario 2, 250 sensor nodes and in scenario 3, 500 sensor nodes have been

From the results obtained for scenario 1, our EERC approach shows performance improvement when compared to the classic LEACH protocol. For 100 nodes, there is 23.85% increase

CH among the nodes in the cluster.

152 Wireless Sensor Networks - Insights and Innovations

overheads and decrease in energy consumption.

Figure 3. Recursive clustering in wireless sensor networks.

4.3. Performance evaluation of EERC

considered.

Similarity Measure is the metric that is employed in this approach for clustering the sensor nodes from a temporal and spatial perspective. The sensor nodes of the same neighborhood produce similar data. Similarly, the data generated from the same sensor node may exhibit similarity to considerable extend except exceptional scenarios of an application. Also, the data that is generated from the same sensor node on successive timeslots in the period of observation may reveal similarity. The redundancy of the data generation and aggregation is effectively controlled through this technique which considerably contributes to the energy consumption in sensor networks.

The component of similarity measure amidst the data sensed from sensor nodes opens the door to devise an effective sleep schedule and save energy. This idea is capitalized in the proposed adaptive distributed clustering algorithm (ADCA).

## 5.2. Phases in adaptive distributed clustering algorithm

It employs two major phases: a cluster formation phase and an adaptive sleep duty cycle phase. In the cluster formation phase, the data generation rate and the similarity between data series are analyzed by the sink. Based on estimation, the nodes are grouped into various clusters. In each cluster, the cluster heads are selected based on the connectivity and residual energy.

In practical scenarios, the clusters may not be in equal size. Based on the similarity measure of the time series, the clustering is done in this approach. By using the similarity measure, the degree of spatial correlation can be calculated. Generally, for two locations with high spatial correlation, their corresponding time series are associated with a high similarity measure. Hence, in a very smooth sub-region, the observed measure has only small changes within the sub-region. In other words, the difference between the observations at any two locations within the sub-region may be very small and hence negligible.

Therefore, without compromising the observation reliability, the working sensor nodes within this sub-region could be sparse. On the other hand, the working sensor nodes should be dense in a fast changing sub-region. The spatial sampling rate has to match the spatial variation of the observed physical incident by setting an appropriate similarity measure threshold value. Hence, the similarity measure threshold value includes a degree of independency. This value can be tuned to balance the trade-off between reliability and energy consumption.

In the sleep duty cycle phase, the data generation rates of cluster members are compared with a minimum threshold level. The nodes which have rates lower than the threshold level are cumulatively allotted a sleep duty cycle for a predefined period. The sleep/wakeup schedule is informed to every sensor node. The scheduling is done on a fair and distributed manner to regulate energy consumption.

Every cluster head collects the data from its members and checks for the similarity in the received data. If it encounters a significant level of change, it reports to the sink along with the data. The sink then performs reclustering or rescheduling of sleep duty cycle, if necessary. Thus, a part of the workload of the sink in periodical checking of the nodes is shared by the cluster heads.

The data values of two sensor nodes ni and nj are said to be similar, if

$$mt\_i = mt\_j\tag{9}$$

where mti and mtj are the magnitudes of the values of ni and nj

$$\mathbf{D}\_{\mathbf{\dot{j}}} < \mathbf{D}\_{\mathbf{\dot{h}}} \tag{10}$$

where Dij is the distance between ni and nj and DTh is the distance threshold.

$$
\delta\_{\dot{i}\dot{j}} < \delta\_{\text{min}\nu} \; \delta\_{\text{R}} \; := \text{ABS}(\mathbf{R}\_{\dot{j}} - \mathbf{R}\_{\dot{i}}) \tag{11}
$$

where Ri and Rj are the sending rates of ni and nj , respectively.

Sending rates are calculated in Eqs. (12) and (13).

$$\mathbf{R}\_{\mathbf{i}} = \mathbf{N} \mathbf{P}\_{\mathbf{i}} / \mathbf{T} \tag{12}$$

$$\mathbf{R}\_{\mathbf{j}} = \mathbf{N} \mathbf{P}\_{\mathbf{j}} / \mathbf{T} \tag{13}$$

here, NPi is the number of packets sent by sensor node i in a time period T.

δij is the absolute difference of sending rates of ni and nj

δmin is the minimum threshold value for δR.

The two nodes can be represented as points in three dimensions. Node i has a set of coordinates (mti, Di and Ri) and Node j has coordinates (mtj , Dj and Rj ).Therefore the Similarity Distance between the nodes ni and nj is given by the Eq. (14).

Modern Clustering Techniques in Wireless Sensor Networks http://dx.doi.org/10.5772/intechopen.70382 155

$$\text{SM}\_{\vec{\eta}} = \sqrt{\sum\_{k=1}^{n} \left(\mathbf{x}\_{\vec{\eta}k} - \mathbf{x}\_{\vec{\eta}k}\right)^{2}}, \ \mathbf{n} = \mathbf{3}. \tag{14}$$

where, xi1 = mti , xi2 = Di, xi3 = Ri and xj1 = mtj , xj2 = Dj , xj3 = Rj .

Here, "n" refers to the number of similarity metrics.

#### 5.3. Cluster formation

Therefore, without compromising the observation reliability, the working sensor nodes within this sub-region could be sparse. On the other hand, the working sensor nodes should be dense in a fast changing sub-region. The spatial sampling rate has to match the spatial variation of the observed physical incident by setting an appropriate similarity measure threshold value. Hence, the similarity measure threshold value includes a degree of independency. This value can be tuned to balance the trade-off between reliability and energy

In the sleep duty cycle phase, the data generation rates of cluster members are compared with a minimum threshold level. The nodes which have rates lower than the threshold level are cumulatively allotted a sleep duty cycle for a predefined period. The sleep/wakeup schedule is informed to every sensor node. The scheduling is done on a fair and distributed manner to

Every cluster head collects the data from its members and checks for the similarity in the received data. If it encounters a significant level of change, it reports to the sink along with the data. The sink then performs reclustering or rescheduling of sleep duty cycle, if necessary. Thus, a part of the workload of the sink in periodical checking of the nodes is shared by the

mti ¼ mtj (9)

Dij < DTh (10)

Ri ¼ NPi=T (12)

Rj ¼ NPj=T (13)

, Dj and Rj

(11)

).Therefore the Similarity

The data values of two sensor nodes ni and nj are said to be similar, if

where Dij is the distance between ni and nj and DTh is the distance threshold.

here, NPi is the number of packets sent by sensor node i in a time period T.

δij < δmin, δ<sup>R</sup> ¼ ABS Rj � Ri

The two nodes can be represented as points in three dimensions. Node i has a set of coordi-

, respectively.

where mti and mtj are the magnitudes of the values of ni and nj

where Ri and Rj are the sending rates of ni and nj

Sending rates are calculated in Eqs. (12) and (13).

δij is the absolute difference of sending rates of ni and nj

nates (mti, Di and Ri) and Node j has coordinates (mtj

Distance between the nodes ni and nj is given by the Eq. (14).

δmin is the minimum threshold value for δR.

consumption.

cluster heads.

regulate energy consumption.

154 Wireless Sensor Networks - Insights and Innovations

In our proposed algorithm, the clustering problem is represented as a clique-covering problem. A graph G is created such that each sensor node is a vertex in the graph. An edge (u,v) between nodes u and v is drawn if SM (u,v) > SMTh.

A cluster is observed as a clique in this problem. A greedy algorithm is used to heuristically find the cliques. Until all vertices are covered, the search starts from the vertex with the largest node degree. The output of this algorithm is a set of cliques that cover all vertices.

#### 5.4. Performance evaluation of ADCA

The performance of the proposed ADCA algorithm has been compared against a contemporary clustering algorithm named energy-efficient distributed clustering (EEDC) algorithm [10]. Number of sensor nodes has been varied from 20 to 100 in a simulation area of 500 � 500 m2 . The results obtained show the performance improvement gained by ADCA with respect to EEDC in terms of energy consumption (25%), delay (18%) and delivery ratio (20%) from the mean measurements of multiple runs.

Thus the effective sleep duty cycle considerably reduces the energy consumption in wireless sensor networks ensuring that the overhead and delay are not increased under such scenarios.

## 6. Summary

Wireless Sensor networks are more sophisticated in their requirements and to provide clustering solutions for them requires adequate knowledge on the nature of the applications, capacity limitations of sensor nodes, the tradeoff among the expected performance parameters and the limitations of emerging technologies.

This chapter has outlined four modern clustering approaches (EAFCA, EEHC, EERC and ADCA) designed for wireless sensor networks, each from a distinguishable perspective. The simulation results obtained for the proposed clustering approaches are encouraging. This will kindle researchers to explore further in this area. The journey of research in this field has crossed significant milestones, yet it has been left with many open-ended issues and unexplored roads due to the presence of inherent trade-offs among the performance factors and dynamic needs of sensor applications. The performance of a wireless sensor network can further be explored through holistic approaches invented or inherited from modern technological advancements.

## Author details

I.S. Akila<sup>1</sup> \*, S.V. Manisekaran<sup>2</sup> and R. Venkatesan<sup>3</sup>


## References


## **Fuzzy Adaptive Setpoint Weighting Controller for WirelessHART Networked Control Systems** Fuzzy Adaptive Setpoint Weighting Controller for WirelessHART Networked Control Systems

DOI: 10.5772/intechopen.70179

Sabo Miya Hassan, Rosdiazli Ibrahim, Nordin Saad, Vijanth Sagayan Asirvadam, Kishore Bingi and Tran Duc Chung Sabo Miya Hassan, Rosdiazli Ibrahim, Nordin Saad, Vijanth Sagayan Asirvadam,

Additional information is available at the end of the chapter Kishore Bingi and Tran Duc Chung

http://dx.doi.org/10.5772/intechopen.70179 Additional information is available at the end of the chapter

#### Abstract

Author details

156 Wireless Sensor Networks - Insights and Innovations

\*, S.V. Manisekaran<sup>2</sup> and R. Venkatesan<sup>3</sup>

\*Address all correspondence to: akila\_subramaniam@yahoo.co.in

3 CSE Department, PSG College of Technology, Coimbatore, India

Computer Communications. 2007;30(14–15):2826-2841

Computing and Communication. 2016;4(4):548-558

Networks. 2012;10(7):1469-1481

on System Sciences (HICSS); Wailea Maui, Hawaii; 2000. pp. 10-19

1 ECE Department, Coimbatore Institute of Technology, Coimbatore, India

[1] Abbasi AA, Younis M. A survey on clustering algorithms for wireless sensor networks.

[2] Heinzelman WB, Chandrakasan AP, Balakrishnan H. Energy-efficient communication protocol for wireless microsensor networks. Proceedings of the International Conference

[3] Taheri H, Neamatollahi P, Younis OM, Naghibzadeh S, Yaghmaee MH. An energy-aware distributed clustering protocol in wireless sensor networks using fuzzy logic. Ad Hoc

[4] Yanagihara K, Taketsugu J, Fukui K, Fukunaga S, Hara S, Kitayama K. EACLE: Energyaware clustering scheme with transmission power control for sensor networks. Wireless

[5] Zhang XF, Yin C. Energy harvesting and information transmission protocol in sensors

[6] Akila IS, Venkatesan R. A cognitive multi-hop clustering approach for wireless sensor networks. Wireless Personal Communications. 2016;90(2):729-747. DOI: 10.1007/s11277-

[7] Akila IS, Venkatesan R. An efficient energy harvesting assisted clustering scheme for wireless sensor networks. International Journal on Recent and Innovation Trends in

[8] Akila IS, Subaselvi S. Energy efficient recursive clustering approach for wireless sensor networks. International Journal of Electronics, Electrical and Computational System.

[9] Manisekaran SV, Venkatesan R. An adaptive distributed power efficient clustering algorithm for wireless sensor networks. . American Journal of Scientific Research. 2010;20:50-63

[10] Liu C, Wu K, Pei J. An energy-efficient data collection framework for wireless sensor networks by exploiting spatiotemporal correlation. IEEE Transaction on Parallel and

Distributed Systems. 2007;18(7):1010-1023. DOI: 10.1109/TPDS.2007.1046

Personal Communications. 2007;40(3):401-415. DOI: 10.1007/s11277-006-9199-2

networks. Journal of Sensors. 2016;2016:1-5. DOI: 10.1155/2016/9364716

2 IT Department, Anna University Regional Campus, Coimbatore, India

I.S. Akila<sup>1</sup>

References

016-3200-5

2017;6(5):121-130

Gain range limitation of conventional proportional-integral-derivative (PID) controllers has made them unsuitable for application in a delayed environment. These controllers are also not suitable for use in a Wireless Highway Addressable Remote Transducer (WirelessHART) protocol networked control setup. This is due to stochastic networkinduced delay and uncertainties such as packet dropout. The use of setpoint weighting strategy has been proposed to improve the performance of the PID in such environments. However, the stochastic delay still makes it difficult to achieve optimal performance. This chapter proposes an adaptation to the setpoint weighting technique. The proposed approach will be used to adapt the setpoint weighting structure to variation in WirelessHART network-induced delay through fuzzy inference. Result comparison of the proposed approach with both setpoint weighting and proportional-integral (PI) control strategy shows improved setpoint tracking and load regulation. For the first-, second- and third-order systems considered, analysis of the results in the time domain shows that in terms of overshoot, undershoot, rise time, and settling times, the proposed approach outperforms both the setpoint weighting and the PI controller. The approach also shows faster recovery from disturbance effect.

Keywords: setpoint weighting, fuzzy adaptation, WirelessHART, PID, wireless sensor networks

## 1. Introduction

Recent advances in wireless technology have prompted researchers to look into its application for industrial process monitoring and control. However, this attempt was hindered by lack of an open and interoperable industrial standard [1–4]. This changed with the coming on board of

© The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons © 2017 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and eproduction in any medium, provided the original work is properly cited.

standards such as WirelessHART, Wireless Networks for Industrial Automation-Process Automation (WIA-PA) and International Society of Automation (ISA) wireless (ISA100.11a). Of these three standards, the WirelessHART has upper hand since it is based on the well-known Highway Addressable Remote Transducer (HART) protocol that is already established with millions of HART-enabled devices already installed worldwide [5–7]. The WirelessHART standard protocol is based on the Open Systems Interconnection model (OSI model) as shown in Figure 1.

The WirelessHART standard adopted a modified version of the physical layer of the IEEE 802.15.4-2006 and operates on the 2.4-GHz industrial, scientific and medical (ISM) radio frequency band. The signals are transmitted over this frequency using 15 channels spaced 5 MHz apart. The time division multiple access (TDMA) method is used for communication whereby packets are sent using 10 ms time slots arranged in the form of superframe. Each superframe thus consists of trains of 10 ms time slots (Figure 2). To avoid interference of other networks and multi-path fading, the standard adopts the strategy of channel hopping between its 15 channels [5, 8]. The standard is secured using the industry standard AES-128 ciphers and keys. The mesh topology of the standard makes it highly reliable, self-organizing and self-healing. In addition to the host computer, a typical WirelessHART network consists of at least a gateway, network manager and field devices as shown in Figure 3.

In spite of the advantages of reduced cabling, improved reliability, scalability and many more offered by wireless technology such as WirelessHART, its application for control is still faced with the challenges of network-induced stochastic delays and uncertainties such as packet

Figure 1. WirelessHART protocol based on OSI layers.

Fuzzy Adaptive Setpoint Weighting Controller for WirelessHART Networked Control Systems http://dx.doi.org/10.5772/intechopen.70179 159

Figure 3. Typical WirelessHART network.

standards such as WirelessHART, Wireless Networks for Industrial Automation-Process Automation (WIA-PA) and International Society of Automation (ISA) wireless (ISA100.11a). Of these three standards, the WirelessHART has upper hand since it is based on the well-known Highway Addressable Remote Transducer (HART) protocol that is already established with millions of HART-enabled devices already installed worldwide [5–7]. The WirelessHART standard protocol is based on the Open Systems Interconnection model (OSI model) as shown in Figure 1.

The WirelessHART standard adopted a modified version of the physical layer of the IEEE 802.15.4-2006 and operates on the 2.4-GHz industrial, scientific and medical (ISM) radio frequency band. The signals are transmitted over this frequency using 15 channels spaced 5 MHz apart. The time division multiple access (TDMA) method is used for communication whereby packets are sent using 10 ms time slots arranged in the form of superframe. Each superframe thus consists of trains of 10 ms time slots (Figure 2). To avoid interference of other networks and multi-path fading, the standard adopts the strategy of channel hopping between its 15 channels [5, 8]. The standard is secured using the industry standard AES-128 ciphers and keys. The mesh topology of the standard makes it highly reliable, self-organizing and self-healing. In addition to the host computer, a typical WirelessHART network consists of at least a gateway,

In spite of the advantages of reduced cabling, improved reliability, scalability and many more offered by wireless technology such as WirelessHART, its application for control is still faced with the challenges of network-induced stochastic delays and uncertainties such as packet

network manager and field devices as shown in Figure 3.

158 Wireless Sensor Networks - Insights and Innovations

Figure 1. WirelessHART protocol based on OSI layers.

dropout. This is as a result of the use of wireless transmitters in the network, which transmit signals aperiodically [9, 10].

From the control perfective, the most common controllers used in the industry are the PID controllers. These controllers are, however, inadequate to be used in a delayed environment [11]. This is because long delays cause oscillation in the response of the system controlled with PID. Furthermore, the PID is limited in gain range, which makes it difficult to adapt to the stochastic nature of the delays in the WirelessHART environment [12]. In an attempt to improve on the performance of the PID in a delayed environment, a setpoint weighting structure was proposed in Ref. [11]. This was later adopted in our work reported in Ref. [13]. The design allows for two degree of freedom control, where both setpoint tracking and good load regulation are achieved. However, if the variability of the network delay is high or if the plant to be controlled is of higher order, the setpoint weighting strategy fails to give optimal performance. Thus, this chapter proposes the adaptation of the setpoint weighting control strategy to the stochastic delay through fuzzy inference system. Fuzzy gain tuning has been an effective way to tune parameters of a controller online with respect to parameter changes. It has been applied recently to tune PID controller for multiple input multiple output (MIMO) systems [14], continuous stirred-tank reactor (CSTR) systems [15], maximum power point tracking in a photovoltaic system [16], load frequency control [17, 18] and many other control applications [19–22].

Among the key advantages of the proposed approach is that although the model of the process to be controlled may be required for the design, it is however not mandatory. Furthermore, in the design, original PID feedback configuration is retained; thus, no modification of the existing structure is required. Finally, the gain range of the PID is significantly extended while achieving robust performance even with external disturbances.

The reminder of this chapter is organized as follows: in section 2, the methodology for the delay measurement is presented, while section 3 gives the design of fuzzy adaptation scheme. The results are presented and discussed in section 4, while in section 5 conclusion is drawn.

## 2. WirelessHART network delay measurement

WirelessHART network delay is measured using Dust Networks DC9007A SmartMesh starter kits produced by Linear Technology. The experimental schematic is shown in Figure 4. The experimental setup consists of a host computer, LTP5903CEN-WHR WirelessHART network manager/Gateway and DC9003-C Eterna WirelessHART motes. As seen from the schematic, the host computer is connected to the gateway through RJ-45 cable, while communication between the gateway and the motes is achieved wirelessly. In this setup, each mote is assumed to be connected to a process plant. Thus, to measure the upstream delay from gateway to the mote tu, and the downstream delay from mote to the gateway td,

Figure 4. WirelessHART network delay measurement schematic.

two-step procedures are involved. First the delay is obtained in the gateway by executing command exec getLatency MACaddress in gateway, where MACaddress is the MAC address of the node in the gateway [13]. Secondly, this delay information is obtained in MATLAB from gateway through the use of Secure Shell (SSH2) software. This is achieved by establishing a secured communication between MATLAB in host and the gateway. The SSH2 command used for this purpose is ssh2\_config ('IP address,' 'userName,' 'password'). The complete procedure is shown in Figure 5.

Figure 5. Procedure for delay measurement.

This is because long delays cause oscillation in the response of the system controlled with PID. Furthermore, the PID is limited in gain range, which makes it difficult to adapt to the stochastic nature of the delays in the WirelessHART environment [12]. In an attempt to improve on the performance of the PID in a delayed environment, a setpoint weighting structure was proposed in Ref. [11]. This was later adopted in our work reported in Ref. [13]. The design allows for two degree of freedom control, where both setpoint tracking and good load regulation are achieved. However, if the variability of the network delay is high or if the plant to be controlled is of higher order, the setpoint weighting strategy fails to give optimal performance. Thus, this chapter proposes the adaptation of the setpoint weighting control strategy to the stochastic delay through fuzzy inference system. Fuzzy gain tuning has been an effective way to tune parameters of a controller online with respect to parameter changes. It has been applied recently to tune PID controller for multiple input multiple output (MIMO) systems [14], continuous stirred-tank reactor (CSTR) systems [15], maximum power point tracking in a photovoltaic system [16], load

Among the key advantages of the proposed approach is that although the model of the process to be controlled may be required for the design, it is however not mandatory. Furthermore, in the design, original PID feedback configuration is retained; thus, no modification of the existing structure is required. Finally, the gain range of the PID is significantly extended while

The reminder of this chapter is organized as follows: in section 2, the methodology for the delay measurement is presented, while section 3 gives the design of fuzzy adaptation scheme. The results are presented and discussed in section 4, while in section 5 conclusion is drawn.

WirelessHART network delay is measured using Dust Networks DC9007A SmartMesh starter kits produced by Linear Technology. The experimental schematic is shown in Figure 4. The experimental setup consists of a host computer, LTP5903CEN-WHR WirelessHART network manager/Gateway and DC9003-C Eterna WirelessHART motes. As seen from the schematic, the host computer is connected to the gateway through RJ-45 cable, while communication between the gateway and the motes is achieved wirelessly. In this setup, each mote is assumed to be connected to a process plant. Thus, to measure the upstream delay from gateway to the mote tu, and the downstream delay from mote to the gateway td,

frequency control [17, 18] and many other control applications [19–22].

160 Wireless Sensor Networks - Insights and Innovations

achieving robust performance even with external disturbances.

2. WirelessHART network delay measurement

Figure 4. WirelessHART network delay measurement schematic.

## 3. Fuzzy adaptive setpoint weighting structure for WirelessHART system (FASW)

This section details the complete design procedure for the fuzzy adaptive setpoint weighting (FASW) control strategy. To do this, the setpoint weighting (SW) structure will first be designed. Then, the fuzzy adaptation will be incorporated to form the FASW structure.

#### 3.1. Setpoint weighting structure

Considering the plant GðsÞ of Eq. (1) in a WirelessHART environment, the typical setpoint weighting strategy for the system as reported in Ref. [13] is shown in Figure 6.

Figure 6. WirelessHART network setpoint weighting structure.

$$G(s) = P(s)e^{-\tau\_p s} = \frac{K\_p}{1 + sT}e^{-\tau\_p s} \tag{1}$$

where Kp, T and τ<sup>p</sup> are the plant gain, time constant and dead-time respectively.

From Figure 6, the closed-loop transfer function from yðsÞ to rðsÞ is given as

$$\frac{y(s)}{r(s)} = \frac{\mathcal{C}(s)P(s)e^{-(\tau\_{ca} + \tau\_p)s}}{1 + \mathcal{C}(s)P(s)e^{-(\tau\_{ca} + \tau\_{bc} + \tau\_p)s}}f\_r(s) \tag{2}$$

where τca and τsc are controller to actuator delay and sensor to controller delay, respectively. In this work, τca ¼ td and τsc ¼ tu.

If τ<sup>1</sup> ¼ τca þ τ<sup>p</sup> and τ<sup>2</sup> ¼ τca þ τsc þ τp, then Eq. (2) becomes

$$\frac{y(s)}{r(s)} = \frac{\mathcal{C}(s)P(s)e^{-\tau\_1 s}}{1 + \mathcal{C}(s)P(s)e^{-\tau\_2 s}}f\_r(s) \tag{3}$$

As reported in our earlier work in Ref. [13], the general setpoint weighting function f <sup>r</sup>ðsÞ is given in the following equation

$$f\_r(\mathbf{s}) = \mathbf{G}\_r(\mathbf{s}) + \check{\mathbf{G}}\_{yr}(\mathbf{s})(e^{-\bar{\mathbf{r}}\cdot\mathbf{s}} - \mathbf{G}\_r(\mathbf{s})) \tag{4}$$

where <sup>G</sup>~yr is the desired closed-loop response, Grðs<sup>Þ</sup> is the feedforward gain enhancement term, and ~τ is the delay estimate. Thus, using Eq. (4) in Eq. (3), we have

$$\frac{y(\mathbf{s})}{r(\mathbf{s})} = \frac{\hat{\mathbf{G}}\_{yr}(\mathbf{s})e^{-\tau\_{1}s}(\mathbf{G}\_{r}(\mathbf{s}) - \mathbf{G}\_{r}(\mathbf{s})\hat{\mathbf{G}}\_{yr}(\mathbf{s}) + \hat{\mathbf{G}}\_{yr}(\mathbf{s})e^{-\tau\_{3}s})}{\mathbf{G}\_{r}(\mathbf{s}) - \mathbf{G}\_{r}(\mathbf{s})\hat{\mathbf{G}}\_{yr}(\mathbf{s}) + \hat{\mathbf{G}}\_{yr}(\mathbf{s})e^{-\tau\_{2}s}}\tag{5}$$

where <sup>G</sup>^ yrðsÞ ¼ GrðsÞCðsÞPðs<sup>Þ</sup> 1þGrðsÞCðsÞPðsÞ .

Under the conditions <sup>τ</sup><sup>~</sup> <sup>¼</sup> <sup>τ</sup>2, <sup>G</sup>^ yrðsÞ ¼ <sup>G</sup>~yrðsÞ, and after pole-zero cancellation, Eq. (5) reduces to

Fuzzy Adaptive Setpoint Weighting Controller for WirelessHART Networked Control Systems http://dx.doi.org/10.5772/intechopen.70179 163

$$\frac{\partial \mathbf{y}(\mathbf{s})}{\partial r(\mathbf{s})} = \hat{G}\_{yr}(\mathbf{s})e^{-\mathbf{r}\_1\mathbf{s}} \tag{6}$$

This indicates that Eq. (6) has decoupled the delay term from the desired closed-loop response <sup>G</sup>^ yrðsÞ. Thus, the implementation of setpoint weighting function <sup>f</sup> <sup>r</sup>ðs<sup>Þ</sup> is shown in Figure 7.

#### 3.2. Design procedures for SW function

To design the proposed fuzzy adaptation scheme, we will first design the setpoint weighting function as follows:

First, the controller CðsÞ is a PI controller given by

$$\mathbf{C}(\mathbf{s}) = \mathbf{K}\_{\mathbb{C}} \left( \mathbf{1} + \frac{1}{T\_i \mathbf{s}} \right) \tag{7}$$

where the proportional gain is related to the system parameters as KC <sup>¼</sup> <sup>0</sup>:5<sup>T</sup> Kpτ<sup>2</sup> and the controller time constant as Ti ¼ T.

If <sup>C</sup>ðs<sup>Þ</sup> is expressed as Acðs<sup>Þ</sup> Bcðs<sup>Þ</sup> , then the feedforward gain enhancement term Grðs<sup>Þ</sup> of <sup>f</sup> <sup>r</sup>ðs<sup>Þ</sup> is designed as follows

$$G\_r(\mathbf{s}) = \frac{\mathbf{K}\mathbf{C}(\mathbf{s})^{-1}\mathbf{P}(\mathbf{s})^{-1}}{B\_t(\mathbf{s})} \tag{8}$$

where K is a tunable gain.

GðsÞ ¼ PðsÞe

From Figure 6, the closed-loop transfer function from yðsÞ to rðsÞ is given as

yðsÞ

Figure 6. WirelessHART network setpoint weighting structure.

162 Wireless Sensor Networks - Insights and Innovations

If τ<sup>1</sup> ¼ τca þ τ<sup>p</sup> and τ<sup>2</sup> ¼ τca þ τsc þ τp, then Eq. (2) becomes

yðsÞ

term, and ~τ is the delay estimate. Thus, using Eq. (4) in Eq. (3), we have

<sup>r</sup>ðs<sup>Þ</sup> <sup>¼</sup> <sup>G</sup>^ yrðsÞe�τ1<sup>s</sup>

.

this work, τca ¼ td and τsc ¼ tu.

given in the following equation

where <sup>G</sup>^ yrðsÞ ¼ GrðsÞCðsÞPðs<sup>Þ</sup>

yðsÞ

1þGrðsÞCðsÞPðsÞ

where Kp, T and τ<sup>p</sup> are the plant gain, time constant and dead-time respectively.

<sup>r</sup>ðs<sup>Þ</sup> <sup>¼</sup> <sup>C</sup>ðsÞPðsÞe�ðτcaþτpÞ<sup>s</sup>

<sup>r</sup>ðs<sup>Þ</sup> <sup>¼</sup> <sup>C</sup>ðsÞPðsÞe�τ1<sup>s</sup>

<sup>f</sup> <sup>r</sup>ðsÞ ¼ GrðsÞ þ <sup>G</sup>~yrðsÞð<sup>e</sup>

where τca and τsc are controller to actuator delay and sensor to controller delay, respectively. In

As reported in our earlier work in Ref. [13], the general setpoint weighting function f <sup>r</sup>ðsÞ is

where <sup>G</sup>~yr is the desired closed-loop response, Grðs<sup>Þ</sup> is the feedforward gain enhancement

Under the conditions <sup>τ</sup><sup>~</sup> <sup>¼</sup> <sup>τ</sup>2, <sup>G</sup>^ yrðsÞ ¼ <sup>G</sup>~yrðsÞ, and after pole-zero cancellation, Eq. (5) reduces to

<sup>ð</sup>GrðsÞ � GrðsÞG~yrðsÞ þ <sup>G</sup>~yrðsÞe�~τ<sup>s</sup>

�τps <sup>¼</sup> Kp

<sup>1</sup> <sup>þ</sup> sT <sup>e</sup>

<sup>1</sup> <sup>þ</sup> <sup>C</sup>ðsÞPðsÞe�ðτcaþτscþτpÞ<sup>s</sup> <sup>f</sup> <sup>r</sup>ðs<sup>Þ</sup> (2)

<sup>1</sup> <sup>þ</sup> <sup>C</sup>ðsÞPðsÞe�τ2<sup>s</sup> <sup>f</sup> <sup>r</sup>ðs<sup>Þ</sup> (3)

�τ~<sup>s</sup> � GrðsÞÞ (4)

Þ

GrðsÞ � GrðsÞG^ yrðsÞ þ <sup>G</sup>^ yrðsÞe�τ2<sup>s</sup> (5)

�τps (1)

Figure 7. Implementation of setpoint weighting structure.

It should be noted that GrðsÞ can be selected simply as K if there is no much information about the system to be controlled.

The desired closed-loop function is thus designed using the following relationship

$$
\hat{G}\_{yr}(\mathbf{s}) = \frac{1}{B\_c(\mathbf{s})/K + 1} \tag{9}
$$

#### 3.3. Fuzzy adaptation mechanism

If the setpoint weighting function f <sup>r</sup>ðsÞ is observed, it can be seen that the terms that depend on the estimate of both the plant dead-time and the network stochastic delay are the gain enhancement term Grðs<sup>Þ</sup> and the delay estimate term <sup>e</sup>�s~<sup>τ</sup> . Thus, in this work, we will use fuzzy adaption mechanism to adjust these parameters accordingly to ensure smooth setpoint tracking and good load regulation. The proposed adaptation mechanism is shown in Figure 8.

The inputs of the supervisor (fuzzy) are the error (e) and its change Δe. The adaptation on f <sup>r</sup>ðsÞ is aiming to correct the system evolution while acting on the control law. During on line operation of the controller, the fuzzy system allows for adaptation of the parameters of the SW function. The change in SW parameters ΔK and Δτ is tuned at each sampling time by using fuzzy adaptation as earlier shown in the figure. The respective ranges of the inputs and outputs of fuzzy tuner are as follows:

$$e, \ \Delta e \in [-2, 2]$$

$$\Delta K \in [-2, 2], \ \Delta \tau \in [0, 2]$$

The range is selected based on the information obtained from the variation of the WirelessHART network delay.

Figure 8. Fuzzy adaptive setpoint weighting structure.

In this proposed fuzzy adaption method, the control rules are developed with the error (e) and change in error (Δe) as a premise and the change in gain (ΔK) and change in delay (Δτ) as consequent of each rule. An example of the tuning rule is given as

IF e is NB and Δe is NB, then ΔK is NVB and Δτ is Z.

It should be noted that GrðsÞ can be selected simply as K if there is no much information about

The desired closed-loop function is thus designed using the following relationship

<sup>G</sup>^ yrðsÞ ¼ <sup>1</sup>

If the setpoint weighting function f <sup>r</sup>ðsÞ is observed, it can be seen that the terms that depend on the estimate of both the plant dead-time and the network stochastic delay are the gain enhancement term Grðs<sup>Þ</sup> and the delay estimate term <sup>e</sup>�s~<sup>τ</sup> . Thus, in this work, we will use fuzzy adaption mechanism to adjust these parameters accordingly to ensure smooth setpoint tracking and good load regulation. The proposed adaptation mechanism is shown in Figure 8.

The inputs of the supervisor (fuzzy) are the error (e) and its change Δe. The adaptation on f <sup>r</sup>ðsÞ is aiming to correct the system evolution while acting on the control law. During on line operation of the controller, the fuzzy system allows for adaptation of the parameters of the SW function. The change in SW parameters ΔK and Δτ is tuned at each sampling time by using fuzzy adaptation as earlier shown in the figure. The respective ranges of the inputs and

e, Δe ∈½�2, 2� ΔK ∈ ½�2, 2�, Δτ∈½0, 2� The range is selected based on the information obtained from the variation of the Wire-

BcðsÞ=<sup>K</sup> <sup>þ</sup> <sup>1</sup> (9)

the system to be controlled.

164 Wireless Sensor Networks - Insights and Innovations

3.3. Fuzzy adaptation mechanism

outputs of fuzzy tuner are as follows:

Figure 8. Fuzzy adaptive setpoint weighting structure.

lessHART network delay.

To achieve smooth adaption, five Gaussian membership functions for input variables and nine Gaussian memberships for output variables have been chosen as shown in Figure 9.

The linguistic descriptions of the input membership functions in the figure are Negative Big (NB), Negative Small (NS), Zero (Z), Positive Small (PS), and Positive Big (PB). The output membership functions of ΔK are Negative Very Big (NVB), Negative Big (NB), Negative Medium (NM), Negative Small (NS), Zero (Z), Positive Small (PS), Positive Medium (PM), Positive Big (PB), and Positive Very Big (PVB). Similarly, the linguistic descriptions for the output membership functions of Δτ are Zero (Z), Very Small (VS), Small (S), Small Medium (SM), Medium (M), Small Big (SB), Medium Big (MB), Big (B), and Very Big (VB).

The 25 fuzzy rules are given in Table 1. The table is generated based on the rule given above. As seen from the table, the first argument of the output represents ΔK, while the second

Figure 9. Fuzzy membership functions.


Table 1. Fuzzy rule table.

Figure 10. Fuzzy rule surface.

argument represents Δτ, i.e., ðΔK, ΔτÞ. The respective rule surfaces for the two outputs based on Table 1 are given in Figure 10.

Fuzzification is achieved using the intersection minimum operation given as follows

$$
\mu\_{A \cap B}(\mathbf{x}, \mathbf{y}) = \min(\mu\_A(\mathbf{x}, \mathbf{y}), \mu\_B(\mathbf{x}, \mathbf{y})) \tag{10}
$$

where A and B are input fuzzy sets (i.e., e and Δe). The values for these inputs are calculated at each sampling time as

$$e(t) = r(t) - y(t) \tag{11}$$

$$
\Delta e = \Delta e(t) - \Delta e(t-1) \tag{12}
$$

For defuzzification, the commonly used centroid method is selected for finding the crisp value of the output. The centroid method is given as:

$$\mu\_o = \frac{\sum\_{i=1}^{R} c\_i \mu\_i}{\sum\_{i=1}^{R} \mu\_i} \tag{13}$$

where


## 4. Results and discussions

This section will present and discuss the results of the proposed approach. In this chapter, three plant models representing first, second and third orders plus dead-time systems are considered. The transfer functions for these models are given in Eqs. (14), (15) and (16), respectively. The parameters of the various controllers used are shown in Table 2. In the table, KC<sup>1</sup> is the controller gain used for the design of the SW controllers, while KC<sup>2</sup> is the proportional gain of the PI controller given in Eq. (7). KC<sup>1</sup> is selected as between 80 and 90% of KC2. The profile and statistical information for the experimental WirelessHART network delay are also given in Figure 11 and Table 3, respectively. Here, the variation in especially upstream delay is observed.

$$P\_1 = \frac{1}{1+2s}e^{-4s} \tag{14}$$

$$P\_2 = \frac{1}{\left(s+1\right)^2} e^{-4s} \tag{15}$$

$$P\_3 = \frac{1}{\left(s+1\right)^3} e^{-5s} \tag{16}$$

#### 4.1. First-order plant

argument represents Δτ, i.e., ðΔK, ΔτÞ. The respective rule surfaces for the two outputs based

e\Δe NB NS Z PS PB NB (NVB, Z) (NB, VS) (NM, S) (NS, SM) (Z, M) NS (NB, VS) (NM, S) (NS, SM) (Z, M) (PS, SB) Z (NM, S) (NS, SM) (Z, M) (PS, SB) (PM, MB) PS (NS, SM) (Z, M) (PS, SB) (PM, MB) (PB, B) PB (Z, M) (PS, SB) (PM, MB) (PB, B) (PVB, VB)

where A and B are input fuzzy sets (i.e., e and Δe). The values for these inputs are calculated at

For defuzzification, the commonly used centroid method is selected for finding the crisp value

X<sup>R</sup> <sup>i</sup>¼<sup>1</sup> ci<sup>μ</sup> X

R <sup>i</sup>¼<sup>1</sup> <sup>μ</sup><sup>i</sup> i

μ<sup>o</sup> ¼

μ<sup>A</sup> <sup>∩</sup> <sup>B</sup>ðx, yÞ ¼ minðμAðx, yÞ, μBðx, yÞÞ (10)

eðtÞ ¼ rðtÞ � yðtÞ (11)

(13)

Δe ¼ ΔeðtÞ � Δeðt � 1Þ (12)

Fuzzification is achieved using the intersection minimum operation given as follows

on Table 1 are given in Figure 10.

of the output. The centroid method is given as:

each sampling time as

Figure 10. Fuzzy rule surface.

Table 1. Fuzzy rule table.

166 Wireless Sensor Networks - Insights and Innovations

The setpoint tracking and disturbance rejection response for P<sup>1</sup> with various controller configurations are given in Figure 12. From the figure, it can be seen that the setpoint tracking ability and disturbance rejection capability of the two setpoint weighted controllers SW and FASW are


Table 2. Controller parameters.

Figure 11. Network delay profile.


Table 3. Network delay statistics.

better than those of the PI controller. The numerical comparison assessed with respect to rise time (Tr), settling time before and after disturbance (Ts<sup>1</sup> and Ts2), overshoot (%OS), and integral time absolute error (ITAE) is given in Table 4. From the table, it is observed that the FASW produced less overshoot of 0.0284% compared to the respective 0.1938 and 4.1582% of SW and PI controllers, while the rise time and settling times of SW are shorter at 4.5980, 19.0756 and 185.5723 s, respectively, than those of FASW and PI.

It is worth noting that the initial control actions of SW and FASW are at 100%, while those of PI are at around 5%. This is due to the improvement of the setpoint weighting ability of the first two controllers.

Fuzzy Adaptive Setpoint Weighting Controller for WirelessHART Networked Control Systems http://dx.doi.org/10.5772/intechopen.70179 169

Figure 12. Response of first-order plant to load disturbance.


Table 4. Performance of first-order plant.

To further evaluate the performance of the controllers, the plant is simulated to a variable setpoint signal and the result is shown in Figure 13. From the responses, it can be seen that during setpoint change both setpoint weighted controllers, i.e., FASW and SW, outperformed the PI controller.

#### 4.2. Second-order plant

better than those of the PI controller. The numerical comparison assessed with respect to rise time (Tr), settling time before and after disturbance (Ts<sup>1</sup> and Ts2), overshoot (%OS), and integral time absolute error (ITAE) is given in Table 4. From the table, it is observed that the FASW produced less overshoot of 0.0284% compared to the respective 0.1938 and 4.1582% of SW and PI controllers, while the rise time and settling times of SW are shorter at 4.5980, 19.0756

Delay type Min Max Mean Standard deviation

Upstream (s) 1.2140 2.0840 1.5734 0.2170 Downstream (s) 1.280 1.280 1.280 0.000

It is worth noting that the initial control actions of SW and FASW are at 100%, while those of PI are at around 5%. This is due to the improvement of the setpoint weighting ability of the first

and 185.5723 s, respectively, than those of FASW and PI.

two controllers.

Figure 11. Network delay profile.

168 Wireless Sensor Networks - Insights and Innovations

Table 3. Network delay statistics.

In a similar way to the first-order plant, the comparison of closed-loop response of this system for setpoint tracking and disturbance rejection with various controllers is shown in Figure 14

Figure 13. Response of first-order plant to changing setpoint.

Figure 14. Response of second-order plant to load disturbance.


Table 5. Performance of second-order plant.

Figure 13. Response of first-order plant to changing setpoint.

170 Wireless Sensor Networks - Insights and Innovations

Figure 14. Response of second-order plant to load disturbance.

and Table 5. From the figure, it is clearly seen that the FASW configuration achieved best tracking and disturbance rejection performance with least overshoot of 0.0286% compared to the 6.1605 and 7.3542% of the SW and PI, respectively. Furthermore, this configuration has the shortest rise and settling times for both before and after disturbance. The initial control signal of both SW and FASW is around 80% while that of the PI is around 10%. Furthermore, the comparison of variable setpoint tracking ability with various controllers is shown in Figure 15. From the responses, just as observed in the first-order plant, the tracking performance of FASW is better than that of SW and PI in terms of overshoot and undershoot during setpoint change.

Figure 15. Response of second-order plant to changing setpoint.

#### 4.3. Third-order plant

In a similar fashion to the earlier two plant models, the comparison of closed-loop response of the third-order system for setpoint tracking and disturbance rejection with various controllers is shown in Figure 16 and Table 6. From both the figure and the table, it is clearly seen that the FASW configuration achieved best tracking and disturbance rejection performance with least overshoot 1.8137% as compared to the 9.3315 and 8.9940% of the SW and PI controllers, respectively. In addition, the proposed configuration has the shortest rise time of around 4.8 s compared to around 7.1 and 13.5 s of the SW and PI controllers. The settling times both before and after disturbance follow the same pattern. The two setpoint weighting configurations SW

Figure 16. Response of third-order plant to load disturbance.


Table 6. Performance of third-order plant.

Fuzzy Adaptive Setpoint Weighting Controller for WirelessHART Networked Control Systems http://dx.doi.org/10.5772/intechopen.70179 173

Figure 17. Response of third-order plant to changing setpoint.

and FASW as observed from the control signals are more aggressive than the PI controller at the beginning: starting at around 50% each.

The comparison of variable setpoint tracking ability with various controllers is shown in Figure 17. From the responses, it is seen that the tracking performance of FASW outperforms those of SW and PI. This is due to the adaptation ability of the FASW controller.

#### 5. Conclusion

4.3. Third-order plant

172 Wireless Sensor Networks - Insights and Innovations

Figure 16. Response of third-order plant to load disturbance.

Table 6. Performance of third-order plant.

In a similar fashion to the earlier two plant models, the comparison of closed-loop response of the third-order system for setpoint tracking and disturbance rejection with various controllers is shown in Figure 16 and Table 6. From both the figure and the table, it is clearly seen that the FASW configuration achieved best tracking and disturbance rejection performance with least overshoot 1.8137% as compared to the 9.3315 and 8.9940% of the SW and PI controllers, respectively. In addition, the proposed configuration has the shortest rise time of around 4.8 s compared to around 7.1 and 13.5 s of the SW and PI controllers. The settling times both before and after disturbance follow the same pattern. The two setpoint weighting configurations SW

Tr Ts<sup>1</sup> Ts<sup>2</sup> %OS ITAE

FASW 3.8653 11.8789 186.2306 0.0286 28.4034 SW 4.1330 37.1544 205.6308 6.1605 30.0163 PI 14.1246 49.2130 205.8253 7.3542 36.7180 This chapter has presented an adaptation mechanism using fuzzy inference system for setpoint weighting controller designed for WirelessHART networked control environment. The adaptation mechanism adjusts the parameters of the setpoint weighting function at each sampling time. Result shows that the proposed approach is able to adapt the controller to variation in network delay. In comparison with ordinary PI controller and fixed setpoint weighting function, the adaptive mechanism has enabled significant improvement of the time domain performance of all the three plants considered. This is even more noticeable in the second- and third-order plants. Future work will focus on the implementation of the approach on a physical plant.

## Author details

Sabo Miya Hassan, Rosdiazli Ibrahim\*, Nordin Saad, Vijanth Sagayan Asirvadam, Kishore Bingi and Tran Duc Chung

\*Address all correspondence to: rosdiazli@utp.edu.my

Department of Electrical and Electronic Engineering, Universiti Teknologi PETRONAS, Perak, Malaysia

## References


[12] Lee TH, Tan KK, Tang KZ. Deadtime compensation via setpoint variation. In: Proceedings of 2009 Asian Control Conference; August 27, 2009; IEEE; 2009. pp. 512-517

Author details

Malaysia

References

Kishore Bingi and Tran Duc Chung

174 Wireless Sensor Networks - Insights and Innovations

2009; IEEE; 2009. pp. 1-9

2013. pp. 185

pp. 1186-1191

\*Address all correspondence to: rosdiazli@utp.edu.my

Sabo Miya Hassan, Rosdiazli Ibrahim\*, Nordin Saad, Vijanth Sagayan Asirvadam,

Department of Electrical and Electronic Engineering, Universiti Teknologi PETRONAS, Perak,

[1] Zheng M, Liang W, Yu H, Xiao Y. Performance analysis of the industrial wireless networks standard: WIA-PA. Mobile Networks and Applications. 2015;22(1):1-12

[2] Petersen S, Carlsen S. WirelessHART versus ISA100. 11a: The format war hits the factory

[3] Petersen S, Carlsen S. Performance evaluation of WirelessHART for factory automation. In: Proceedings of Emerging Technologies & Factory Automation; ETFA; September 22,

[4] Miya HS, Ibrahim RB, Saad NB, Asirvadam VS, Chung TD. WirelessHART process control with Smith predictor compensator. Advanced Science Letters. 2016;22(10):2676-2680

[6] Olenewa J. Guide to Wireless Communication. Cengage Learning; Boston, MA, USA

[7] Hassan SM, Ibrahim R, Bingi K, Chung TD, Saad N. Application of wireless technology for control: A WirelessHART perspective. Procedia Computer Science. 2017;105:240-247

[8] Chung TD, Ibrahim RB, Asirvadam VS, Saad NB, Hassan SM. Adopting EWMA filter on a fast sampling wired link contention in WirelessHART control system. IEEE Transac-

[9] Blevins TL. PID advances in industrial control. IFAC Proceedings Volumes. 2012;45(3):23-28 [10] Blevins T, Chen D, Han S, Nixon M, Wojsznis W. Process control over real-time wireless sensor and actuator networks. In: Proceedings of High Performance Computing and Communications (HPCC), 2015 IEEE 7th International Symposium on Cyberspace Safety and Security (CSS), 2015 IEEE 12th International Conference on Embedded Software and Systems (ICESS), 2015 IEEE 17th International Conference; August 24, 2015; IEEE; 2015.

[11] Tan KK, Tang KZ, Su Y, Lee TH, Hang CC. Deadtime compensation via setpoint varia-

[5] Chen D, Nixon M, Mok A. Why WirelessHART. US: Springer; 2010. pp. 195-199

tions on Instrumentation and Measurement. 2016;65(4):836-845

tion. Journal of Process Control. 2010;20(7):848-859

floor. IEEE Industrial Electronics Magazine. 2011;5(4):23-34


**Provisional chapter**

## **A Hybrid Sink Repositioning Technique for Data Gathering in Wireless Sensor Networks Gathering in Wireless Sensor Networks**

**A Hybrid Sink Repositioning Technique for Data** 

DOI: 10.5772/intechopen.70335

## Prerana Shrivastava Prerana Shrivastava Additional information is available at the end of the chapter

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/intechopen.70335

#### **Abstract**

Wireless sensor network (WSN) is a wireless network that consists of spatially distributed autonomous devices using sensors to cooperatively investigate physical or environmental conditions. WSN has a hundreds or thousands of nodes that can communicate with each other and pass data from one node to another. Energy can be supplied to sensor nodes by batteries only and they are configured in a harsh environment in which the batteries cannot be charged or recharged simply. Sensor nodes can be randomly installed and they autonomously organize themselves into a communication network. The main constraint in wireless sensor networks is limited energy supply at the sensor nodes so it is important to deploy the sink at a position with respect to the specific area which is the area of interest; which would result in minimization of energy consumption. Sink repositioning is very important in modern day wireless sensor network since repositioning the sink at regular interval of time can balance the traffic load thereby decreasing the failure rate of the real time packets. More attention needs to be given on the Sink repositioning methods in order to increase the efficiency of the network. Existing work on sink repositioning techniques in wireless sensor networks consider only static and mobile sink. Not much importance is given to the hybrid sink deployment techniques. Multiple sink deployment and sink mobility can be considered to perform sink repositioning. Precise information of the area being monitored is needed to offer an ideal solution by the sink deployment method but this method is not a realistic often. To reallocate the sink, its odd pattern of energy must be considered. In this chapter a hybrid sink repositioning technique is developed for wireless sensor network where static and mobile sinks are used to gather the data from the sensor nodes. The nodes with low residual energy and high data generation rate are categorized as urgent and the nodes with high residual energy and low data generation rate are categorized as non-urgent. Static sink located within the center of the network collects the data from the urgent nodes. A relay is selected for each urgent sensor based on their residual energy. The urgent sensor sends their data to the static sink through these relay. Mobile sink collects the data from the non-urgent sensors. The performance of the proposed technique is compared with mobile base station placement scheme mainly based on the performance according to the metrics such as average end-to-end delay, drop, average packet delivery ratio and average energy consumption.

© 2016 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. © 2017 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Through the simulation results it is observed that the proposed hybrid sink repositioning technique reduces the energy hold problem and minimizes the buffer overflow problem thereby elongating the sensor network lifetime.

**Keywords:** wireless sensor networks, sink repositioning, energy efficiency, hybrid technique, network lifetime

## **1. Introduction**

## **1.1. Wireless sensor network**

A wireless sensor network (WSN) is a discrete network comprising of numerous wireless nodes referred to as sensors, which are deployed in order to perform the designated specific tasks like monitoring the surroundings and measuring the physical parameters such as temperature, pressure, humidity, etc. Since the location of an individual sensor cannot be preplanned or predetermined, these networks must have the potential to self-organize themselves. A wider geographic area can be covered by efficiently networking a large number of sensors thereby resulting in precise, dependable and robust networks. Wireless sensor networks are responsible to gauge, record, process and transfer the information to the destination node within the network zone using the assigned communication routes. Each sensor deployed in the network performs the functions like sensing the environment, processing the sensed data and communicating with the neighboring sensors. The sensor nodes have limited sensing range, processing power and energy levels.

The performance and efficiency of any wireless sensor network depends on the computational power, battery lifetime, data storage and communication bandwidth which in turn are directly dependent on the available energy levels. A major hurdle in the operation of sensors is the unavailability of an adequate energy. Normally the sensors depend upon their battery for power which in many cases cannot be replaced or recharged. Hence while designing any protocol for such networks, the conservation of the available energy of the sensor must be considered as an important factor. Thus, extending the lifetime of the sensor networks is a major area which is receiving a significant amount of interest from the research communities.

#### *1.1.1. Structure of wireless sensor network*

A basic sensor network consists of a large amount of sensor nodes. Each sensor is made up of small individual microcontroller fitted with sensors in which communication such as radios is used. The components of a sensor node are a sensing unit, a processing unit, a transceiver and a power unit. Generally, the sensor networks can form either a mesh topology or a star topology. Nodes can propagate by routing or flooding. In WSN, each node is assigned a number as its unique address for the purpose of communication. Functionally sensor nodes can be classified into two types. First, the nodes that deal within the network with other nodes and second, the ones which interface with the outside environment which are called as the gateway nodes or the sink nodes. The general structure of the wireless sensor network is shown in **Figure 1**.

As shown in **Figure 1**, the number of sensors is deployed in the geographical extent of the entire network and they will perform their task of sensing, processing, relaying and doing communication. All the information or the data that is sensed by the sensors will be forwarded to the sink node through multi hop relaying from where it will be provided to the end users.

**Figure 1.** Structure of wireless sensor network.

Through the simulation results it is observed that the proposed hybrid sink repositioning technique reduces the energy hold problem and minimizes the buffer overflow problem

**Keywords:** wireless sensor networks, sink repositioning, energy efficiency, hybrid technique,

A wireless sensor network (WSN) is a discrete network comprising of numerous wireless nodes referred to as sensors, which are deployed in order to perform the designated specific tasks like monitoring the surroundings and measuring the physical parameters such as temperature, pressure, humidity, etc. Since the location of an individual sensor cannot be preplanned or predetermined, these networks must have the potential to self-organize themselves. A wider geographic area can be covered by efficiently networking a large number of sensors thereby resulting in precise, dependable and robust networks. Wireless sensor networks are responsible to gauge, record, process and transfer the information to the destination node within the network zone using the assigned communication routes. Each sensor deployed in the network performs the functions like sensing the environment, processing the sensed data and communicating with the neighboring sensors. The sensor nodes have limited

The performance and efficiency of any wireless sensor network depends on the computational power, battery lifetime, data storage and communication bandwidth which in turn are directly dependent on the available energy levels. A major hurdle in the operation of sensors is the unavailability of an adequate energy. Normally the sensors depend upon their battery for power which in many cases cannot be replaced or recharged. Hence while designing any protocol for such networks, the conservation of the available energy of the sensor must be considered as an important factor. Thus, extending the lifetime of the sensor networks is a major area which is receiving a significant amount of interest from the

A basic sensor network consists of a large amount of sensor nodes. Each sensor is made up of small individual microcontroller fitted with sensors in which communication such as radios is used. The components of a sensor node are a sensing unit, a processing unit, a transceiver and a power unit. Generally, the sensor networks can form either a mesh topology or a star topology. Nodes can propagate by routing or flooding. In WSN, each node is assigned a number as its unique address for the purpose of communication. Functionally sensor nodes can be classified into two types. First, the nodes that deal within the network with other

thereby elongating the sensor network lifetime.

sensing range, processing power and energy levels.

network lifetime

178 Wireless Sensor Networks - Insights and Innovations

**1.1. Wireless sensor network**

research communities.

*1.1.1. Structure of wireless sensor network*

**1. Introduction**

## *1.1.2. Types of wireless sensor networks*

Wireless sensor network typically has little or no infrastructure. There are two types of WSNs, namely structured model and an unstructured model. Structured model is deployed in a preplanned manner and it is used only for the network with fewer nodes. It has lower network maintenance and cost. Uncovered regions are not present in this model. Unstructured model is densely deployed in the network. The nodes that are deployed randomly have uncovered regions and are left unattended to perform the task. Maintenance is difficult here.

#### *1.1.3. Characteristics of wireless sensor network*

Some of the salient characteristics of the wireless sensor networks are described below:


## *1.1.4. Major applications of wireless sensor networks*

There are various applications of WSN that require constant monitoring and particular event detecting based on the requirement and features of the system. The applications can be divided into three categories [2] as mentioned in **Table 1**.

The importance of WSN is briefly described below for certain major applications as follows:

• Environmental and agricultural applications

WSNs are useful for the purpose of area monitoring, monitoring water levels as well as the rainfall. It is also used for forest fire & flood monitoring. Agricultural applications include sensing of chemicals, soil condition, irrigation planning, etc.

• Military applications

WSN's various characteristics are extremely useful in the area of enemy movement tracking, enemy intelligence information collection and transmission, surveillance, etc.

• Medical operations

Sensor networks play a critical role in monitoring the physiological readings of patient like blood pressure or pulse, etc. It plays an extremely significant role in post calamity medical relief operations such as earthquakes and floods.

• Heavy industrial monitoring

WSNs help in industrial applications by enabling to track material movement, warehousing, inventory planning and refurbishing, in spite of the harsh field conditions, consequently saving huge costs that are involved in such type of businesses.


**Table 1.** Categories of application.

#### **1.2. Sink repositioning**

• Energy, computation and storage limitations: sensor nodes have limited energy, computation, and storage capabilities. Hence, the energy conservation measures are required in

• Self-configurable: generally, the sensor nodes can be randomly installed and they are capable of establishing a communication network by organizing themselves appropriately. • Unreliable sensor nodes and data redundancy: sensor nodes are prone to physical damages or outages owing to their deployment in harsh or hostile conditions. The sensor nodes that are deployed close to each other play a similar role, in order to accomplish a common sensing task in a given area of concern. This results in building up of redundancy in events of failure. • Application specific: depending upon the application, the design considerations of wireless

• Frequent topology change: in most of the sensor network applications, the sensed data may pass through the various sensor nodes between source and the sink, thereby showing a many-to-one traffic pattern. The sensor node failure, damage, energy depletion, etc. may

There are various applications of WSN that require constant monitoring and particular event detecting based on the requirement and features of the system. The applications can be

The importance of WSN is briefly described below for certain major applications as follows:

WSNs are useful for the purpose of area monitoring, monitoring water levels as well as the rainfall. It is also used for forest fire & flood monitoring. Agricultural applications include

WSN's various characteristics are extremely useful in the area of enemy movement tracking,

Sensor networks play a critical role in monitoring the physiological readings of patient like blood pressure or pulse, etc. It plays an extremely significant role in post calamity medical

WSNs help in industrial applications by enabling to track material movement, warehousing, inventory planning and refurbishing, in spite of the harsh field conditions, consequently saving

enemy intelligence information collection and transmission, surveillance, etc.

order to improve the efficiency and the life of the network.

180 Wireless Sensor Networks - Insights and Innovations

sensor network will vary and may need customization.

force the network topology to change continuously [1].

divided into three categories [2] as mentioned in **Table 1**.

sensing of chemicals, soil condition, irrigation planning, etc.

*1.1.4. Major applications of wireless sensor networks*

• Environmental and agricultural applications

relief operations such as earthquakes and floods.

huge costs that are involved in such type of businesses.

• Military applications

• Medical operations

• Heavy industrial monitoring

In WSN, sinks are bounded with abundant resources and the sensors that generate data are termed as sources. The sources can transmit data to a single or multiple sinks for the purpose of analysis and processing.

In wireless sensor networks, sink repositioning is preferred almost by all applications that involve real time communication. It helps to evenly distribute the traffic and hence minimize the packet loss or the data loss. To carry out sink repositioning, multiple sink deployment and sink mobility can be adopted. Precise information of the area being monitored is needed to offer an ideal solution by repositioning the sink.

#### *1.2.1. Types of sink repositioning*

Sink repositioning can be performed in the following ways.

Multiple sink deployment: in a given geographic area, multiple sinks can be deployed. By deploying multiple sinks in the network, the average number of hops through which the information has to pass through is decreased, since the data will always be sent to the nearest sink. Also by deploying multiple sinks, the load is evenly distributed among all the sinks [3].

Sink mobility: it is extremely advantageous in case of WSNs, if the sinks can move within the network boundaries with an acceptable delay. The mobile sink collects the data from the sensor nodes and also transmits it further. Although this approach results in comparatively higher time lag or latency, it helps in conserving the energy and hence increasing the life span [4].

Deploying multiple mobile sinks: multiple mobile sinks can be deployed in order to collect the data from the sensors in the given network without causing delay and buffer overflow problem. Here the mobile sink will relocate at regular intervals before the sensor's buffer overflows thereby avoiding the buffer overflow problem.

Initially, the research work in the field of wireless sensor networks mainly discussed the issues related to an uneven energy consumption which was leading to the energy hole problem in a sensor network. Generally, all the sensors generate data at a constant bit rate and transmit the data to the static sink through multihop transmission. Therefore the sensors which are closer to sink will die of energy soon, thereby creating an energy hole around the sink. The researchers have proposed an analytical modeling for the energy hole problem and using their model they have discussed the effectiveness of various techniques employed for justifying this problem.

## **2. Problem statement**

During the regular network operation, relocating the sink is very challenging. During the sink's movement, the fundamental issues are when the sink should move, where the sink should move and how the data traffic would be handled when the sink is on the move. In a multi-hop network, finding an optimal location for the sink is very difficult. The difficulty mainly arises due to the following two factors. First, the sink can be moved to an infinite possible position. Secondly, a new multi-hop network topology needs to be established for every solution considered during the search for an optimal location [5].

Since employing the sink requires the precise knowledge of the monitored area, they are not always reasonable, even though the sink deployment can provide optimal solution. When accurate position of sensor is available and when nodes have motion capabilities, controlled deployment or online deployment is possible. The developing graph may have different properties during the online deployment. The basic issue in the sensor deployment is controlling the dynamic graph of mobile sensor networks [6]. The energy-unbalanced problem is another big challenge in sink deployment. Here the sensors that are closer to the sink are likely to consume their energy much faster than the other nodes [7] . When a network consists of multiple clusters, the relocation problem is significantly compounded. The sink cannot choose to move randomly around its cluster to enhance the intra-cluster network operation without considering the potential impact on inter sink connectivity that could impose on its capability to maintain communication with the sink nodes of other clusters [8]. Using the odd pattern of energy depletion, first the relocation of the sink has to be initiated even if it is considered as the most efficient network operation for a given traffic distribution at that time. The sink must make sure that no data is lost, when it is moving [9]. Using mobile sinks for data gathering has the drawback of buffer overflow problem. In other words, the sink has to visit each sensor node before its buffer overflows and this will depend on the speed of the mobile sink. However, it is very difficult to set the optimum speed for the mobile sink, since each sensor node has different buffer sizes and information generation rate. Apart from this problem, the residual energy of the sensors must also be considered, since sensors with low residual energy may deplete their energy before the mobile sink visits them.

## **3. Research issues addressed**

In order to deal with the various issues in case of the wireless sensor network, the main objectives of the research is to design and implement a hybrid sink repositioning technique (HSRT) for data gathering in wireless sensor networks. The main focus has been on devising a technique which draws the benefits of both multiple sinks and sink repositioning, in order to improve the energy efficiency and various other performance metrics of the network. The design aspects of HSRT have been aimed at overcoming the energy hole problem and buffer overflow problem by taking into consideration the residual energy of the sensors that are deployed in the network.

## **4. Research methodology**

sink. The researchers have proposed an analytical modeling for the energy hole problem and using their model they have discussed the effectiveness of various techniques employed for

During the regular network operation, relocating the sink is very challenging. During the sink's movement, the fundamental issues are when the sink should move, where the sink should move and how the data traffic would be handled when the sink is on the move. In a multi-hop network, finding an optimal location for the sink is very difficult. The difficulty mainly arises due to the following two factors. First, the sink can be moved to an infinite possible position. Secondly, a new multi-hop network topology needs to be established for every

Since employing the sink requires the precise knowledge of the monitored area, they are not always reasonable, even though the sink deployment can provide optimal solution. When accurate position of sensor is available and when nodes have motion capabilities, controlled deployment or online deployment is possible. The developing graph may have different properties during the online deployment. The basic issue in the sensor deployment is controlling the dynamic graph of mobile sensor networks [6]. The energy-unbalanced problem is another big challenge in sink deployment. Here the sensors that are closer to the sink are likely to consume their energy much faster than the other nodes [7] . When a network consists of multiple clusters, the relocation problem is significantly compounded. The sink cannot choose to move randomly around its cluster to enhance the intra-cluster network operation without considering the potential impact on inter sink connectivity that could impose on its capability to maintain communication with the sink nodes of other clusters [8]. Using the odd pattern of energy depletion, first the relocation of the sink has to be initiated even if it is considered as the most efficient network operation for a given traffic distribution at that time. The sink must make sure that no data is lost, when it is moving [9]. Using mobile sinks for data gathering has the drawback of buffer overflow problem. In other words, the sink has to visit each sensor node before its buffer overflows and this will depend on the speed of the mobile sink. However, it is very difficult to set the optimum speed for the mobile sink, since each sensor node has different buffer sizes and information generation rate. Apart from this problem, the residual energy of the sensors must also be considered, since sensors with low

solution considered during the search for an optimal location [5].

residual energy may deplete their energy before the mobile sink visits them.

In order to deal with the various issues in case of the wireless sensor network, the main objectives of the research is to design and implement a hybrid sink repositioning technique (HSRT) for data gathering in wireless sensor networks. The main focus has been on devising a technique which draws the benefits of both multiple sinks and sink repositioning, in order to

**3. Research issues addressed**

justifying this problem.

182 Wireless Sensor Networks - Insights and Innovations

**2. Problem statement**

## **4.1. Structure overview of hybrid sink repositioning technique (HSRT)**

In the hybrid sink repositioning technique (HSRT), the sensors are randomly deployed within the geographic extent of the entire network. A single static sink and multiple mobile sinks are deployed in the network. The static sink is deployed at the center of the network. In case of sensors, the overflow of information occurs due to the limited storage capacity. The overflow time of each sensor is computed based upon their storage size and the data generation rate. All the sensors are then allotted a particular group based on their overflow time and location. After this one mobile sink is assigned to each group. Depending upon the data generation rate and residual energy of the sensors, the sensors are classified into two different categories namely urgent and non-urgent sensors. The static sink performs the function of collecting the data from the urgent sensors. A strategy has been devised in order to select and form the set of relay sensors, in such a manner that every individual urgent sensor has at least single relay sensor that is closest to the static sink. The urgent sensors transmits their information through the relay sensors to the final destination which is the static sink. In order to collect the data from the non-urgent sensors, a mobile sink deployment algorithm has been developed which will periodically collect the data from these sensors.

#### **4.2. Sensor node classification**

To explain the concept, a wireless sensor network with "i" number of sensors is considered. The sensor node classification has been done into two groups as urgent sensors and nonurgent sensors based on their residual energy and the data generation rate.

As shown in **Figure 2**, Eri is the residual energy of the sensors, DGri is the data generation rate of the sensors, Ert is the minimum threshold value of the residual energy and DGrt is the maximum threshold value of the data generation rate. The sensors are classified as urgent and non-urgent sensors depending on the following two criteria.

If, Eri < Ert and DGri > DGrt then the sensor is treated as urgent sensor.

Else if,

Eri > Ert and DGri < DGrt then the sensor is treated as non-urgent sensor.

Thus a sensor having low residual energy and high data generation rate is categorized as urgent sensor and the sensor having high residual energy and low data generation rate is categorized as non-urgent sensors.

**Figure 2.** Classification of the sensors.

#### **4.3. Positioning relay sensors near the static sink**

For effective network operation and optimum performance, a two layer network is considered in a sensing field as shown in **Figure 3** wherein the relays and the static sink form the upper layer whereas the urgent sensors form the bottom layer.

**Figure 3.** Formation of set of relay sensors.

Let,

N be the static sink,

S = {S1 , S2 , S3 … Sn} be the set of urgent sensors,

V = {V1 , V2, V3 … Vk } be the set of non-urgent sensors and.

R = {R<sup>1</sup> , R<sup>2</sup> , R<sup>3</sup> … R<sup>e</sup> } be the set of relay sensors.

In the given sensing field, the sensors are densely deployed whereas the relays are sparsely deployed. Data gathering is done by the joint co-operation of both sensors and the relays. A relay sensor is connected to the static sink in the upper layer of the network otherwise it is unconnected. Initially the set of the urgent sensors and the relays is not known as shown in **Figure 4**.

The main concern is to make use of the relay sensors having high residual amount of energy, in order to forward the information that is sensed by the urgent sensors to the static sink. A set of primary relays which are nearest connected relays to the urgent sensors S is determined. Let this set of relay sensors be denoted by M (SRne).The urgent sensors directs their data to M (SRne) and then M (SRne) relays this sensory data to the static sink N. In each interval the set of the relay sensors keeps on changing.

Now a set H(m<sup>i</sup> ) is created such that,

$$\mathbf{H}(\mathbf{m}\_{\downarrow}) = \{ \mathbf{S}\_{\text{n}} \, | \, \text{M}(\text{SR}\_{\text{nv}}) = \mathbf{m}\_{\downarrow} \} \tag{1}$$

where H(m<sup>i</sup> ) is the set of all the urgent sensors attended by m<sup>i</sup> . Each mi will cover a set of all the relay sensors M(SRne) for all the urgent sensors Sn in different groups as depicted in **Figure 3**.

**Figure 4.** Two layer network.

**Figure 3.** Formation of set of relay sensors.

**4.3. Positioning relay sensors near the static sink**

**Figure 2.** Classification of the sensors.

184 Wireless Sensor Networks - Insights and Innovations

upper layer whereas the urgent sensors form the bottom layer.

Urgent Sensors nsors

RE i < RE th and GR i > GR th

For effective network operation and optimum performance, a two layer network is considered in a sensing field as shown in **Figure 3** wherein the relays and the static sink form the

Sensors

Non-Urgent Sensors

> RE i > RE th and GR i < GR th

#### **4.4. Mobile sink deployment algorithm**

A mobile sink deployment algorithm is developed in order to collect the data periodically from the non-urgent sensors. The main objective of this algorithm is to ensure that when the mobile sink is on the move it must travel minimum distance and at the same time perform maximum data collection.

All the mobile sinks, will first of all, identify those non-urgent sensors which are directly transmitting their information to them and at the same time the distance between these sensors and the mobile sink is less than the particular threshold value of the transmission distance. This is done because each mobile sink has its own capability of till what distance it can move while relocating. So a threshold value of the transmission distance for the mobile sink is selected. A set CJ is created where CJ denotes the set of the id's of those non-urgent sensors which are sending their information directly to the mobile sinks.

$$\mathbf{C}\_{\mathbf{j}} = \left\{ \mathbf{I} = D\_{\mathbf{j}}^{(0)} \le \mathbf{T} \mathbf{D}\_{\text{th}'} \text{ I} \in \mathbf{V} \right\} \tag{2}$$

Once the set CJ is created, the mobile sink will wait for a particular duration during which each non-urgent sensor from CJ will transmit minimum one data packet to the mobile sink. The header of the data packet holds the ids of those sensors which are transmitting their information through these non-urgent sensors. As soon as the mobile sink receives the data packet from the non-urgent sensors from the set CJ , it records the ids of such sensors which are sending their data through these non-urgent sensors. Finally the mobile sink is able to identify the number of such sensors which are transmitting their own information through K, where, K ∈ C<sup>J</sup> .

In order to reduce the mean distance between the non-urgent sensors and the mobile sink, the position of the distant sensors needs to be estimated. For this a set Zk is created such that,

$$Z\_{\mathbf{k}} \text{ = } \# \{ \mathbf{l} ; \mathbf{k} \equiv \min \mathbf{D}\_{\mathbf{l}}^{\mathbf{u}} \} , \mathbf{k} \in \text{route}\_{\mathbf{u} \mathbf{l}} \} \tag{3}$$

Where Zk is the set of the number of those distant sensors that transmit their information through the non-urgent sensors to the mobile sink and at the same time, the distance between them and the non-urgent sensors is minimum. Here routeIK is the set of id's of the sensors on the route from sensor I to the non-urgent sensor k.

Once the mobile sink has identified that there are Z<sup>k</sup> sensors communicating through nonurgent sensors k, the next task is to find the optimal position for the mobile sink. For this, the resultant route vector is used. The resultant route vector for sink j is approximated as,

$$\mathbf{R}\,\mathbf{V}\_{\rangle} = \frac{\sum\_{\mathbf{k}\in\mathbb{Q}\_{\times}^{\times}} \mathbb{U}\_{\mathbf{k}}{}^{\otimes}.\,\mathbf{Z}\_{\mathbf{k}}}{\sum\_{\mathbf{k}}\mathbb{Z}\_{\mathbf{k}}},\mathbf{j} = \text{ 1...K} \tag{4}$$

where, RV<sup>j</sup> is the Resultant route vector; U<sup>k</sup> (j) is the unit vector from mobile sink j to the nonurgent sensor k; and Zk is the set of number of distant sensors communicating through k.

A Hybrid Sink Repositioning Technique for Data Gathering in Wireless Sensor Networks http://dx.doi.org/10.5772/intechopen.70335 187

**Figure 5.** System flowchart of the HSRT Algorithm.

**4.4. Mobile sink deployment algorithm**

186 Wireless Sensor Networks - Insights and Innovations

is created where CJ

CJ = {I = *D*<sup>I</sup>

packet from the non-urgent sensors from the set CJ

Zk =#{I:k = min D<sup>I</sup>

the route from sensor I to the non-urgent sensor k.

R Vj =

Once the mobile sink has identified that there are Z<sup>k</sup>

is the Resultant route vector; U<sup>k</sup>

each non-urgent sensor from CJ

.

which are sending their information directly to the mobile sinks.

maximum data collection.

is selected. A set CJ

Once the set CJ

K, where, K ∈ C<sup>J</sup>

Where Zk

where, RV<sup>j</sup>

urgent sensor k; and Zk

A mobile sink deployment algorithm is developed in order to collect the data periodically from the non-urgent sensors. The main objective of this algorithm is to ensure that when the mobile sink is on the move it must travel minimum distance and at the same time perform

All the mobile sinks, will first of all, identify those non-urgent sensors which are directly transmitting their information to them and at the same time the distance between these sensors and the mobile sink is less than the particular threshold value of the transmission distance. This is done because each mobile sink has its own capability of till what distance it can move while relocating. So a threshold value of the transmission distance for the mobile sink

The header of the data packet holds the ids of those sensors which are transmitting their information through these non-urgent sensors. As soon as the mobile sink receives the data

are sending their data through these non-urgent sensors. Finally the mobile sink is able to identify the number of such sensors which are transmitting their own information through

In order to reduce the mean distance between the non-urgent sensors and the mobile sink,

through the non-urgent sensors to the mobile sink and at the same time, the distance between them and the non-urgent sensors is minimum. Here routeIK is the set of id's of the sensors on

urgent sensors k, the next task is to find the optimal position for the mobile sink. For this, the

resultant route vector is used. The resultant route vector for sink j is approximated as,

<sup>∑</sup><sup>k</sup>∈Qj Uk (j) . Zk \_\_\_\_\_\_\_\_\_\_\_ ∑Zk

(k)

is the set of the number of those distant sensors that transmit their information

is the set of number of distant sensors communicating through k.

the position of the distant sensors needs to be estimated. For this a set Zk

is created, the mobile sink will wait for a particular duration during which

denotes the set of the id's of those non-urgent sensors

will transmit minimum one data packet to the mobile sink.

(J) < TDth,  I ∈ V} (2)

, it records the ids of such sensors which

, k ∈ routeIK} (3)

sensors communicating through non-

, j = 1...K (4)

(j) is the unit vector from mobile sink j to the non-

is created such that,

If the magnitude of the resultant route vector is less than a particular threshold value, then the mobile sink rests at its current position. On the other hand, if the resultant route vector is greater than a particular threshold value, then the mobile sink will reposition itself to a new location PJ + RV<sup>j</sup> . Stmax, where, PJ is the current position of the mobile sink and St is the maximum probable value of the stride that can be achieved by the mobile sink.

The process is repeated and the iteration continues, if the mobile sinks are moving for collecting the data from the non-urgent sensors. If all the mobile sinks have come to a standstill, then the mobile sink deployment algorithm terminates.

## **4.5. System flowchart of hybrid sink repositioning technique (HSRT)**

The overall system flowchart of the hybrid sink repositioning technique (HSRT) that is designed for the purpose of data gathering in case of wireless sensor networks is depicted in **Figure 5**.

## **5. Research outputs and results**

## **5.1. Simulation model and parameters**

The implementation and the simulation of the hybrid sink repositioning technique (HSRT) are done by using the Network Simulator ns 2.32. A bounded region of 1000 × 1000 m2 is considered in which the sensors are deployed using a rectangle distribution. The power levels are assigned to the sensors in such a way that their communication and sensing range is 250 m. In the simulation, the maximum data that can be supported by the communication media is fixed to 2 Mbps. The traffic generator used is the constant bit rate. The medium access control layer protocol used for the wireless local area network is the distributed coordination function of IEEE 802.11.

**Table 2** depicts the various network parameters and their values which are assigned in the simulation model.


**Table 2.** Network parameters.

All the energy values have been selected based upon the energy model of ns2.32. Energy model represents the level of the energy in the sensors like the initial energy, idle energy and the usage of the energy for every packet it transmits and receives. The TCL script has been written for the HSRT. The NAM file is executed from the TCL script and it displays the network visualization of the HSRT. The NAM output which gives us the network visualization of HSRT is shown in **Figure 6**.

A single static sink as indicated by red color is deployed within the center of the network. The various sensors that are deployed in the network are assigned to a particular group. The multiple mobile sink as indicated by blue color are deployed in the network of the HSRT wherein each group is allotted one mobile sink, which will relocate itself inside the group that has been assigned to it, around every specific interval of time in order to collect the data from the non-urgent sensors. After running and executing the simulation, the mobile sink repositions itself to a new optimal location which is computed by the HSRT Algorithm, in order to collect the data from the non-urgent sensors of that particular group, as shown in **Figure 7**.

The number of sensors deployed in the network is increased and correspondingly the NAM output is observed before and after the sink repositioning by employing the HSRT, as shown in **Figures 8** and **9** respectively.

**Figure 6.** NAM output of HSRT.

If the magnitude of the resultant route vector is less than a particular threshold value, then the mobile sink rests at its current position. On the other hand, if the resultant route vector is greater than a particular threshold value, then the mobile sink will reposition itself to a new

The process is repeated and the iteration continues, if the mobile sinks are moving for collecting the data from the non-urgent sensors. If all the mobile sinks have come to a standstill, then

The overall system flowchart of the hybrid sink repositioning technique (HSRT) that is designed for the purpose of data gathering in case of wireless sensor networks is depicted in **Figure 5**.

The implementation and the simulation of the hybrid sink repositioning technique (HSRT) are

in which the sensors are deployed using a rectangle distribution. The power levels are assigned to the sensors in such a way that their communication and sensing range is 250 m. In the simulation, the maximum data that can be supported by the communication media is fixed to 2 Mbps. The traffic generator used is the constant bit rate. The medium access control layer protocol used for the wireless local area network is the distributed coordination function of IEEE 802.11. **Table 2** depicts the various network parameters and their values which are assigned in the

done by using the Network Simulator ns 2.32. A bounded region of 1000 × 1000 m2

mum probable value of the stride that can be achieved by the mobile sink.

**4.5. System flowchart of hybrid sink repositioning technique (HSRT)**

is the current position of the mobile sink and St is the maxi-

is considered

location PJ + RV<sup>j</sup>

simulation model.

**Table 2.** Network parameters.

. Stmax, where, PJ

188 Wireless Sensor Networks - Insights and Innovations

the mobile sink deployment algorithm terminates.

**5. Research outputs and results**

**5.1. Simulation model and parameters**

Area size 1000 × 1000 m2 MAC IEEE 802.11 Traffic source CBR Routing protocol AODV Simulation time 50 s Packet size 500 Bits Idle power 0.035 W Transmit power 0.660 W Receive power 0.395 W Initial energy 10.1 J

Number of sensors 20, 40, 60, 80, 100

Rate 50, 100, 150, 200 and 250 kb

**Figure 7.** NAM output of HSRT after sink repositioning.

**Figure 8.** NAM output of HSRT with increased number of sensors.

#### **5.2. Simulation results**

The evaluation of the performance of the hybrid sink repositioning technique that is designed is done based on the four performance metrics of any wireless sensor network. These performance metrics are the average energy consumption, end to end delay, average packet drop and packet delivery ratio. All these parameters play a vital role in assessment of any designed technique, since the main focus is on data gathering application of the wireless sensor network. A Hybrid Sink Repositioning Technique for Data Gathering in Wireless Sensor Networks http://dx.doi.org/10.5772/intechopen.70335 191

**Figure 9.** NAM output of HSRT after sink repositioning with increased number of sensors.

**Figure 7.** NAM output of HSRT after sink repositioning.

190 Wireless Sensor Networks - Insights and Innovations

**Figure 8.** NAM output of HSRT with increased number of sensors.

The evaluation of the performance of the hybrid sink repositioning technique that is designed is done based on the four performance metrics of any wireless sensor network. These performance metrics are the average energy consumption, end to end delay, average packet drop and packet delivery ratio. All these parameters play a vital role in assessment of any designed technique, since the main focus is on data gathering application of the wireless sensor network.

**5.2. Simulation results**

The performance of HSRT is compared with the existing multiple mobile base station placement scheme (MBSP) [10] for doing the necessary evaluation.

The tracing and monitoring of the simulation is done by running the TCL script which gives the trace values. The analysis of these trace values that has resulted from the simulation is done by making use of the trace data analyzer which is the X-Graph. The X-Graph is called within the OTCL script. The X-Graph will visually display the information of the trace values produced from the simulation.

The effect of HSRT on the various mentioned performance metrics is seen first by varying the number of sensors in the network and then by varying the speed of the mobile sinks.

## *5.2.1. Simulation results obtained by varying the number of sensors and the speed of the mobile sinks*

In order to analyze the scalability of the HSRT, the number of sensors is varied from 20 to 100. The trace values for both HSRT and MBSP are monitored. **Figures 10**–**17** show the graphical representation of the simulation results obtained for various performance metrics by employing both HSRT and MBSP.

**Figure 10** shows the average energy consumption for both the techniques, when the number of sensors is increased. The energy consumption increases almost linearly for the two techniques, when the network size is increased. It is observed that the HSRT consumes less energy when compared to the existing MBSP, since the relays are selected based on their residual energy. Moreover a particular threshold value of the residual energy is set for the sensors and therefore before the sensors completely deplete their energy, the proposed HSRT technique comes into picture and proper strategy as described is implemented which results in the significant amount of the energy saving of the entire network.

**Figure 10.** Sensors vs. average energy consumption.

**Figure 11.** Speed vs. average energy consumption.

**Figure 12.** Sensors vs. end to end delay.

A Hybrid Sink Repositioning Technique for Data Gathering in Wireless Sensor Networks http://dx.doi.org/10.5772/intechopen.70335 193

**Figure 13.** Speed vs. end to end delay.

**Figure 14.** Sensors vs. average drop.

<sup>0</sup> <sup>1</sup> <sup>2</sup> <sup>3</sup> <sup>4</sup> <sup>5</sup> <sup>6</sup> <sup>7</sup> <sup>8</sup> <sup>9</sup> <sup>10</sup> <sup>0</sup>

<sup>0</sup> <sup>10</sup> <sup>20</sup> <sup>30</sup> <sup>40</sup> <sup>50</sup> <sup>60</sup> <sup>70</sup> <sup>80</sup> <sup>90</sup> <sup>100</sup> <sup>0</sup>

Number of Sensors

energy. Moreover a particular threshold value of the residual energy is set for the sensors and therefore before the sensors completely deplete their energy, the proposed HSRT technique comes into picture and proper strategy as described is implemented which results in the sig-

> MBSP HSRT

> > MBSP HSRT

MBSP HSRT

Speed (m/s)

0 10 20 30 40 50 60 70 80 90 100

Number of Sensors

0

0.2

E nd t o E n d D e l a y ( S e c )

0.4

0.6

0.8

nificant amount of the energy saving of the entire network.

A v e r a g e E n e r g y C ons u m p t i o n

**Figure 10.** Sensors vs. average energy consumption.

192 Wireless Sensor Networks - Insights and Innovations

A v e r a g e E n e r g y C o n s u m p t i o n

**Figure 11.** Speed vs. average energy consumption.

**Figure 12.** Sensors vs. end to end delay.

**Figure 15.** Speed vs. average drop.

In order to analyze the mobility of sinks, the speed of the mobile sinks is varied from 2 to 10 m/s. **Figure 11** shows the average energy consumption of both HSRT and MBSP when the speed of the mobile sink is increased. The energy consumption increases linearly as observed from the simulation. Moreover HSRT consumes less energy than MBSP.

**Figure 16.** Sensors vs. packet delivery ratio.

**Figure 17.** Speed vs. packet delivery ratio.

**Figure 12** shows the average end-to-end delay in a scenario of varying number of sensors. When the network size is increased, it increases sink deployment time leading to the increased delay. From **Figure 12**, it is observed that HSRT minimizes the delay when compared with the existing MBSP scheme. In the proposed HSRT, the average number of hops that are involved in the transmission and reception of the data is minimized which also leads to the reduction of the overall end to end delay. Moreover the mobile sink is itself relocating at regular intervals to collect the data from the non-urgent sensors.

**Figure 13** shows the results of average end-to-end delay when the speed of mobile sinks is increased from 2 to 10 m/s. It is observed that the delay increases beyond 0.3 seconds when the speed is above 6 m/s. The proposed HSRT shows a significant amount of improvement in end to end delay as compared to MBSP due to the proper distribution of the traffic load between the sinks as well as the sensors.

**Figure 14** gives the average drop occurred for both the techniques when the number of sensors is increased. The increase in network size results in slight increase in packet drop. It can be seen that HSRT has less packet drop when compared to the existing MBSP, since all kind of losses that results from the energy hole problem and the buffer overflow problem are taken care of while designing the proposed hybrid sink repositioning Technique which results in the minimization of the packet drops.

On the other hand, **Figure 15** gives the results of packet drop for both the techniques when the speed of mobile sink is increased. When the mobile sink moves at higher speed, more buffer overflow will occur thereby resulting in more packet drops. This drawback is overcome in the proposed HSRT. Simulation results indicates that HSRT results in the reduction of the dropping of the data packets with the increasing speed of the mobile sinks.

**Figure 16** shows the corresponding packet delivery ratio by varying the number of sensors. The increase in the network size results in the slight degradation of the delivery ratio. It is observed that HSRT achieves higher packet delivery ratio when compared with the existing MBSP technique. In HSRT the traffic load is evenly distributed among the sinks as well as the sensors. The use of mobile sinks which are relocating at regular intervals also decreases the number of hops. This ensures the enhancement in the delivery of the packets with less drop in the packets. The reduction in the Average Drop of the proposed HSRT gives rise to the improvement of the packet delivery ratio.

**Figure 17** presents the packet delivery ratio when the speed of the mobile sink is increased. Higher the speed of the mobile sink larger will be the packet drops. Hence the delivery ratio decreases. But due to the optimum relocation of the sinks and data flow pattern, HSRT achieves higher delivery ratio than MBSP.

#### *5.2.2. Percentage improvement of HSRT over existing MBSP*

**Table 3** shows the percentage improvement of HSRT when compared to the existing MBSP scheme.


**Table 3.** Percentage improvement of HSRT.

**Figure 12** shows the average end-to-end delay in a scenario of varying number of sensors. When the network size is increased, it increases sink deployment time leading to the increased delay. From **Figure 12**, it is observed that HSRT minimizes the delay when compared with the existing MBSP scheme. In the proposed HSRT, the average number of hops that are involved in the transmission and reception of the data is minimized which also leads to the reduction of the overall end to end delay. Moreover the mobile sink is itself relocating at regular intervals

<sup>0</sup> <sup>1</sup> <sup>2</sup> <sup>3</sup> <sup>4</sup> <sup>5</sup> <sup>6</sup> <sup>7</sup> <sup>8</sup> <sup>9</sup> <sup>10</sup> <sup>0</sup>

<sup>0</sup> <sup>10</sup> <sup>20</sup> <sup>30</sup> <sup>40</sup> <sup>50</sup> <sup>60</sup> <sup>70</sup> <sup>80</sup> <sup>90</sup> <sup>100</sup> <sup>0</sup>

Number of Sensors

MBSP HSRT

MBSP HSRT

Speed ( m/s )

**Figure 13** shows the results of average end-to-end delay when the speed of mobile sinks is increased from 2 to 10 m/s. It is observed that the delay increases beyond 0.3 seconds when the speed is above 6 m/s. The proposed HSRT shows a significant amount of improvement in end to end delay as compared to MBSP due to the proper distribution of the traffic load

**Figure 14** gives the average drop occurred for both the techniques when the number of sensors is increased. The increase in network size results in slight increase in packet drop. It can be seen that HSRT has less packet drop when compared to the existing MBSP, since all kind of losses that results from the energy hole problem and the buffer overflow problem are taken

to collect the data from the non-urgent sensors.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.2 0.4 0.6 0.8 1 1.2 1.4

Packet

Delivery Ratio

P a c k e t D e l i v e r y R a t i o

**Figure 17.** Speed vs. packet delivery ratio.

**Figure 16.** Sensors vs. packet delivery ratio.

194 Wireless Sensor Networks - Insights and Innovations

between the sinks as well as the sensors.

## **6. Research conclusion**

The hybrid sink repositioning technique is less complex and the overheads involved in running the algorithm is less and hence the proposed HSRT technique can be easily implemented in any real time applications like for the purpose of surveillance, military application or any other scenario where efficient data gathering is the prime focus and where each and every event needs to be detected properly.

Moreover the major hurdles directly affecting the performance of wireless sensor networks, namely the energy hole problem and the buffer overflow problem are minimized by the proposed hybrid sink repositioning technique that has been designed and successfully implemented. Through simulation results, it has been observed that employing the HSRT Algorithm enhances the overall functioning of the entire wireless sensor network in terms of the performance metrics namely the average energy consumption, end to end delay, average drop and packet delivery ratio. The improvement in all these performance metrics extends the lifetime as well as the accuracy of the WSN. Moreover HSRT also reduces the complexity involved in repositioning the multiple mobile sinks by employing the mobile sink deployment algorithm at regular intervals efficiently.

The research work presented in this chapter mainly focused on the energy consumption in terms of balancing and saving in order to extend the lifetime of the WSNs. Basically all the mobile sinks should remain active all the time in order to perform the task of data collection efficiently. But, also there are chances that they will remain idle most of the time if the data collection is light. The process of data collection gets affected or comes to a standstill if the mobile sink fails due to some fault. So the future scope or work must aim towards development of some effective visiting schedule and trajectory for the mobile sinks. Moreover it should include techniques for reducing the power consumption of idle sink and recovery of the failed mobile sinks.

## **Author details**

Prerana Shrivastava

Address all correspondence to: prerana01@hotmail.com

Lokmanya Tilak College of Engineering, Navi Mumbai, India

## **References**


[3] Vincze Z, Vida R, Vidacs A. Deploying multiple sinks in multihop wireless sensor networks. Proceedings of IEEE Conference on Pervasive Services. July 2007

**6. Research conclusion**

196 Wireless Sensor Networks - Insights and Innovations

event needs to be detected properly.

ment algorithm at regular intervals efficiently.

Address all correspondence to: prerana01@hotmail.com

Lokmanya Tilak College of Engineering, Navi Mumbai, India

the failed mobile sinks.

**Author details**

Prerana Shrivastava

**References**

The hybrid sink repositioning technique is less complex and the overheads involved in running the algorithm is less and hence the proposed HSRT technique can be easily implemented in any real time applications like for the purpose of surveillance, military application or any other scenario where efficient data gathering is the prime focus and where each and every

Moreover the major hurdles directly affecting the performance of wireless sensor networks, namely the energy hole problem and the buffer overflow problem are minimized by the proposed hybrid sink repositioning technique that has been designed and successfully implemented. Through simulation results, it has been observed that employing the HSRT Algorithm enhances the overall functioning of the entire wireless sensor network in terms of the performance metrics namely the average energy consumption, end to end delay, average drop and packet delivery ratio. The improvement in all these performance metrics extends the lifetime as well as the accuracy of the WSN. Moreover HSRT also reduces the complexity involved in repositioning the multiple mobile sinks by employing the mobile sink deploy-

The research work presented in this chapter mainly focused on the energy consumption in terms of balancing and saving in order to extend the lifetime of the WSNs. Basically all the mobile sinks should remain active all the time in order to perform the task of data collection efficiently. But, also there are chances that they will remain idle most of the time if the data collection is light. The process of data collection gets affected or comes to a standstill if the mobile sink fails due to some fault. So the future scope or work must aim towards development of some effective visiting schedule and trajectory for the mobile sinks. Moreover it should include techniques for reducing the power consumption of idle sink and recovery of

[1] Coleri S, Puri A, Varaiya P. Power efficient system for sensor networks. Proceedings of 8th IEEE International Conference on Computers and Communication. July 2003

[2] Akkaya K, Younis M, Bangad M. Sink repositioning for enhanced performance in wireless sensor networks. Elsevier Journal on Computer Networks. 2005;**49**:512-534


## *Edited by Philip Sallis*

Wireless sensor networks (WSNs) have emerged as a phenomenon of the twenty-first century with numerous kinds of sensor being developed for specific applications. The origins of WSNs can, however, be traced back to the early days of connectivity between computers and their peripherals. Work with distributed sensor networks is evidenced in the literature during the latter part of the 1970s, continuing in functionality increases in the 1980s and 1990s. As a configuration of independent devices in a data communications network, WSNs are now pre-eminent as working solutions to numerous precision data collection situations where software control of instruments and routing protocols are needed. In this book, the authors have chosen a selection of specific topics relating to WSNs: their design, development, implementation and function. Some operating topics are addressed such as power management, data interchange protocols, instrument reliability and system security. Other topics are more application oriented, where particular hardware and software configurations are described to deliver system solutions for specific needs. All are clearly written with considerable detail relating to each of the issues addressed by the authors. Each of the chapters provides a rationale for the topic being covered and some general WSN details where appropriate. The citations used in the chapters are comprehensively referred to, which adds depth to the information being presented.

> ISBN 978-953-51-3561-6 ISBN 978-953-51-4635-3

Wireless Sensor Networks - Insights and Innovations

Wireless Sensor Networks

Insights and Innovations

*Edited by Philip Sallis*

Photo by chombosan / iStock