A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;November 4, 2016;19:8]

Computers and Electrical Engineering 0 0 0 (2016) 1–19

Contents lists available at ScienceDirect

Computers and Electrical Engineering journal homepage: www.elsevier.com/locate/compeleceng

A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networksR Khalid Haseeb a,∗, Kamalrulnizam Abu Bakar a, Abdul Hanan Abdullah a, Adnan Ahmed b, Tasneem Darwish a, Fasee Ullah a a b

Faculty of Computing, Universiti Teknologi Malaysia UTM, 81310 Skudai, Johor, Malaysia Quaid-e-Awam University of Engineering, Science & Technology, Nawabshah 67480, Sindh, Pakistan

a r t i c l e

i n f o

Article history: Received 25 February 2016 Revised 25 October 2016 Accepted 26 October 2016 Available online xxx Keywords: Network lifetime Energy-awareness Clusters formation Fault tolerant Energy de-efficiency Wireless sensor networks

a b s t r a c t In recent decades, cluster-based schemes have emerged as viable solutions for energy conservation problem in wireless sensor networks (WSN). However, most of the existing solutions incur imbalanced energy consumption and high network overheads. In addition, existing cluster-based solutions do not optimize route discovery based on the restricted resources of sensor nodes. Moreover, most of the cluster-based solutions perform periodical re-clustering for load balancing, which results in shortening network lifetime. This research paper presents a Dynamic Energy-aware Fault Tolerant Routing (DEFTR) protocol that exploits uniform-sized network partitioning based on network size and utilizes a multi-facet routing mechanism, which takes into consideration the residual energy, and position and link quality of neighbors. Furthermore, DEFTR not only offers reliable and energy efficient data routing but also supports fault tolerance. Simulation results demonstrate that DEFTR improved the network lifetime by 20.9% and throughput by 35%, also it reduced the delay by 29% and transmission cost by 46% in comparison to the existing work. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction WSN comprises of an enormous number of tiny devices called sensor nodes that are spread in a physical environment for collecting required information. With the integration of information sensing, computation, and wireless communication, the sensor nodes can sense physical status, process the sensed information, and report them to the sink node or Base Station (BS) [1,2]. Due to limited resources, sensor nodes consume their energy rapidly and it is impractical or difficult to refresh their batteries. Consequently, improving energy conservation is the foremost design challenge for any sensor based network. In recent decades, the concept of hierarchical based solutions [3,4] has been widely used in the field of wireless communication for energy efficiency and routing performance. Basically, the concept behind hierarchical schemes is to group the sensor nodes in non-overlapping regions called clusters and each cluster has one leader node called Cluster Head(CH) [5]. Although different clustering schemes have been proposed, most of them generate sub-optimal clusters leading to unnecessary energy consumption and imbalanced load distribution [6,7]. In addition, most of the existing schemes perform non-optimized route discoveries, which disturb the data forwarding operations and may result in frequent route breakages R This paper is for CAEE special section SI-wls7. Reviews processed and recommended for publication to the Editor-in-Chief by Guest Editor Dr. R. Varatharajan. ∗ Corresponding author. E-mail address: [email protected] (K. Haseeb).

http://dx.doi.org/10.1016/j.compeleceng.2016.10.017 0045-7906/© 2016 Elsevier Ltd. All rights reserved.

Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE 2

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

with re-transmissions. Moreover, the majority of the recent schemes support fault tolerance by performing re-clustering of the entire network field at periodic intervals, which acquires additional communication overheads and energy resources. In this paper, a dynamic energy-aware fault tolerant routing protocol is designed to address the limitations of existing energybased routing schemes for WSNs. Unlike the majority of existing solutions, our proposed protocol generates uniformly sized clusters based on network size. By exploiting composite metrics within the bounded regions, appropriate nodes are chosen to be CHs thereby improving network lifetime. Based on a multi-facet strategy, the optimized routing paths are constructed at multi-level and a fault tolerant mechanism is developed, which result in reducing route recovery time and balancing energy consumption. The simulations for evaluating DEFTR protocol are done in Network simulator (NS-2) [8,9]. The simulation results reveal outstanding performance in terms of different evaluation metrics as compared to other state-of-the-art protocols. The remainder of this paper is structured as follows. Section 2 presents the related work. Section 3 highlights the findings of related work and contributions of proposed protocol. Section 4 illustrates the preliminaries and energy model. Section 5 presents the details of our proposed DEFTR protocol. The evaluation and validation of proposed DEFTR protocol are discussed in Section 6. Finally, Section 7 concludes the paper with some future research directions. 2. Related work Sensor nodes of WSNs are characterized by constrained resources.During data gathering and forwarding process, the sensors limited energy has impacts on network lifetime. LEACH [10] is the first dynamic cluster-based routing approach for WSN, which uses a stochastic process. A random number is generated by each node n then it is elected as a CH if the random number is smaller than a threshold Tn as shown in Eq. 1, where P is the required fraction of clusters, r is the current round, G is the set of nodes that have been not selected as CH in last 1/p rounds.



Tn =

P 1−P∗ r mod

0

(

1 P

)

if n ∈ G otherwise

(1)

However, random selection of CHs leads to short network lifetime and sub-optimal clusters formation. To improve the performance of LEACH, different energy efficient schemes are proposed [11,12], where CHs are selected by BS based on the provided information by nodes. Nevertheless, such solutions have a scalability problem and incur further network overheads. The Multi-hop Routing with Low Energy Adaptive Clustering Hierarchy (MR-LEACH) [13] is presented for balancing the energy load and improving data delivery performance. MR-LEACH approach divides the sensors field into numerous layers and each layer is considered as a particular cluster. MR-LEACH initiated the concept of equal clustering, where any node in a particular layer can reach BS in identical hop count. However, it lacks of optimizing the end-to-end routing paths by keeping in view the scarce resources of nodes, thereby results in lower network throughput. Chain-Chain Based Routing Protocol (CCBRP) [14] provides a hybrid scheme that incorporates the characteristics of LEACH and PEGASIS. CCBRP allocates the nodes into different chains in two different phases, which incurs high energy consumption in the large scaled region. In addition, the selection of intermediate nodes during chain construction is non-optimized which significantly decreases network lifetime and requires more re-transmissions. Different unequal cluster-based schemes [8,15,16] are developed to address the issue of a hot spot, which occurs in neighbor clusters closest to BS, as these clusters need to handle additional traffic load coming from the faraway clusters. In such schemes, unequal clusters are generated where smaller size clusters are closer to BS. Nevertheless, utilizing different size of clusters leads to uneven distribution of traffic load among CHs and has scalability problems, as the overall operation is performed in a centralized manner.Recently, fuzzy based clustering protocols are presented [17,18] for the formation of clusters and electing of CHs over the network field. As compared to probabilistic approaches, fuzzy based schemes are more energy efficient and reliable. However, such schemes have slow convergence time and are not feasible for large scale networks. In Tree Based Clustering (TBC) [19], a logical tree is constructed inside each cluster where the associated CH is elected as the root node. TBC utilizes distance factor for the determining of nodes level inside each constructed tree. However, TBC lacks the ability to adapt to changing network conditions such as node’s residual energy and quality of wireless links, which results in frequently route breakages and compromised network lifetime. An energy efficient fault tolerant clustering routing protocol [20] was proposed, which aims to execute run time recovery when CHs are unexpectedly fail. In this protocol, a Distributed Fault tolerant Clustering Routing (DFCR) algorithm is introduced to perform clustering and data routing phases. To handle the failure of CHs during data routing, DFCR searches an additional suitable CH with minimum hop counts. In addition, to deal with fault tolerance, this approach only considered permanent failures of CHs and temporary failure of sensors is overlooked. 3. Contributions of proposed DEFTR protocol Based on the related work, it is revealed that cluster-based schemes have a great impact on saving energy and have higher network scalability. However, most of the existing clustering schemes are based on probabilistic or random selection methods that construct sub-optimal clusters. Moreover, the CH election mechanism incurs over the entire network field, which leads to degrading network lifetime and significantly increasing network overheads. Furthermore, most of existing cluster-based routing schemes do not optimize route discovery according to the restricted resources of sensor nodes and the dynamic nature of wireless communication links. In addition, due to limited and low-powered constraints of sensor Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

[m3Gsc;November 4, 2016;19:8] 3

nodes, most of the existing energy efficient routing schemes are prone to faults because of imbalanced energy consumption. An exhausted node not only disturbs the data delivery performance but it also affects the network lifetime. It is observed that to provide fault tolerance mechanism, most of the existing schemes either exploit re-transmissions or replicate the same message on different routes which results in extra network congestion. Moreover, it is concluded that most of the existing schemes update the role of CHs among nodes on the periodical basis, for the purpose of enhancing their fault tolerance level, which leads to unnecessary overheads. This research paper focuses on developing energy-aware cluster-based fault tolerance routing protocol, which is capable of generating balanced sized clusters and optimized routes. The proposed protocol is suitable for wide range of important applications that require energy efficient, reliable and multi-hop communication. Firstly, DEFTR protocol divides the sensor nodes into distinct clusters based on network size, which places near optimal number of nodes in each cluster and has significant improvement in terms of energy consumption and network lifetime. In addition, the proposed protocol employs the CH election mechanism inside the clusters boundaries by exploiting multiple weighted metrics. Secondly, DEFTR exploits composite routing function for the routing decision, which is based on node’s distance, residual energy, and link quality factors. The composite function determines a limited number of intermediate nodes to take part in route discovery mechanism, which results in reducing communication overheads. The adopted multi-metric strategy of proposed protocol at both intra and inter-clusters communication is to establish shortest, more energy efficient and consistent routing paths, which improves network lifetime. At the end, DEFTR reduces network disruption, which is caused by exhausted or energy de-efficient nodes over the constructed routing paths. In order to balance the energy consumption over the routing paths, DEFTR constructs alternative disjoint routes within each cluster boundary, which can be used when the primary route is broken. Moreover, based on network conditions, DEFTR makes dynamic decisions to replace the exhausted CHs rather than re-clustering the entire network field at periodic intervals, which improves the end-to-end network connectivity. The qualitative analysis of our proposed DEFTR protocol with existing solutions is summarized in Table 1. 4. Preliminaries This section presents the network assumptions and energy model that are adopted by DEFTR protocol. Sensor nodes are randomly deployed in two-dimensional network field and the entire network field comprises of homogenous nodes in terms of resources. After deployment, all nodes remain static and each node knows its location via a positioning algorithm, which is applied only once after deployment. Nodes can amend their transmission power based on receiver distance and sense data at a fixed interval. During network setup, BS is considered the most powerful node and does not have any resources restriction. We considered the energy model that was adopted in [10,20]. Assume that d is a distance between two nodes i and j. The energy consumption during transmitting and receiving k data bits is shown in Eqs. 2 and 3.



ET r (k, d ) =

Eelect ∗ k + k ∗ E f s ∗ d2 , i f d ≤ dt Eelect ∗ k + k ∗ Eamp ∗ d4 , i f d > dt

ERx (k ) = Eelect ∗ k

(2) (3) ∗ d2

Eelect is per bit energy consumption during sending and receiving . The amplifier’s energy consumption is shown by Efs or Efs ∗ d4 which is selected on the basis distance from source to destination. Distance threshold is given by dt , if d ≤ dt then free space model is used where multi-hop fading model is employed. 5. Design of DEFTR protocol This section presents our DEFTR protocol, whereas the details of its components are discussed in sub sequent sections.The design of DEFTR protocol constitutes clusters formation, dynamic routes setup and fault tolerability components. Firstly, clusters formation partitions the entire sensor field into uniformly sized clusters based on network size. In addition, using compound weighted metrics, the appropriate nodes are chosen as CHs that are optimally fit to be the administrative heads within the clusters. Secondly, the dynamic routes setup incorporates multi-facet strategy in routing decisions to achieve shortest, more energy efficient and reliable data forwarding. Finally, the aim of fault tolerance component is to support robust routing by discovering alternative routes when energy-deficient nodes are encountered on active routing path. Moreover, CHs are focal points within clusters regions; as they are exposed to high energy consumption due to their additional liabilities and being frequently involvement in routes re-adjustment process. Therefore, the role of CHs is shifted on demand basis among cluster’s members by exploiting network conditions 5.1. Clusters formation In DEFTR protocol, the process of clusters formation is executed only once in entire network lifetime. Thus, DEFTR protocol reduces network overheads and energy expenses by avoiding frequent re-clustering. Based on nodes density, BS creates Time Division Multiple Access (TDMA)schedules and requests the nodes to localize themselves. BS broadcasts its ID and upon receiving a BS discovery packet, each node makes an entry in its routing table. In addition, nodes increment the hop count field in the packet and further share the BS discovery packet with their neighbors. A node might receive aBS discovery Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

Fault tolerance

Link-Quality

Hop Counts

Trans-mission cost

Through-Put

Routing Metric

Network Lifetime

end-to-end delay

Robust

Distributed Centralized Distributed Distributed Distributed Distributed Distributed Distributed Distributed Distributed Distributed

No No No No No No No No 1-level 1-level k-level

No No No No No No No No No No Yes

1-hop 1-hop k-hop k-hop 1-hop k-hop k-hop k-hop k-hop k-hop Multi level

Moderate High Moderate Moderate Moderate Moderate High Moderate Moderate High Moderate

Low Low Moderate Moderate Low Moderate Moderate Low Moderate Moderate High

Direct Direct Shortest Shortest Direct Shortest Energy, distance Direct Shortest Shortest Optimal Routing

Low Low Moderate Moderate Low Moderate Moderate Low Moderate Moderate High

Moderate Moderate High Moderate Moderate Moderate High Moderate Moderate Moderate Moderate

Limited Limited Moderate Moderate Limited Limited Limited Limited Moderate Moderate High

ARTICLE IN PRESS

Method

LEACH Dynamic/central clustering CCBRP MR-LEACH Unequal leach UHEED TRP TBC DFCR Optimal load distribution Proposed Protocol

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

Protocol

JID: CAEE

4

[m3Gsc;November 4, 2016;19:8]

Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

Table 1 Qualitative analysis of proposed protocol and existing schemes.

ARTICLE IN PRESS

JID: CAEE

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

[m3Gsc;November 4, 2016;19:8] 5

Fig. 1. Impact of number of nodes on clusters formation.

packet from multiple sources, in this case, source node offering least hop count towards BS are favored and recorded in the routing table. Eventually, all nodes learn their route towards BS. Afterwards, each node sends its position information via its upstream node and this step continues till BS has a global knowledge of the entire network field. Then, based on network size (n), BS divides the entire network field into an optimum number of clusters (p) of square sized partitions by using Eq. 4, where k is considered as a squared number.

k=n∗ p

(4)

In the design of DEFTR protocol, the vital functionality is to find out the optimum number of clusters during a network partition. If the number of clusters is non-optimized it offset the benefits of clustering mechanism and leads to inefficient use of energy and computational resources. Fig. 1 illustrates that number of generated clusters is directly proportional to network size and increases linearly as network size increases from 100 to 500 nodes. To commence with clusters formation phase, BS exploits the knowledge of geographical positions (xi , yi ) of known nodes and computes the center point for each partition i. Next, by using node’s position and determined set of centre points, for each virtual partition BS formulates a cluster of the set of nodes which are relatively closer to its centre point. Furthermore, a unique CLUS_ID is assigned to each generated cluster. Such mechanism is sustained until all nodes are collected into non-overlapping k clusters. The foremost aim of this mechanism is to distribute roughly identical number of nodes per cluster and chances of two immediate nodes to be selected as CHs are avoided. Moreover, most of the existing schemes perform CH election in probabilistic or random manner, which results in the construction of sub-optimal clusters and leads to unbalanced energy consumption with compromised network lifetime. A CH is an essential point and over burdened with data traffic, which results in energy hole problems and degrading network performance. The next stage is to initiate the election mechanism for initial CHs within the bounded regions. A minimum subset of nodes participates for CH election, which leads to fewer communication overheads and energy consumption. The proposed clustering component performs CH election mechanism in distributed manner and exploits compound metrics by considering node’s residual energyei , distance dti and densitydi factors. Based on these factors, each node inside a cluster region computes its weighted value. As a result, nodes that optimize the weighted metric are considered as candidates for CHs. Firstly, nodes determine their distance dti towards centroid of clusters by using a Euclidean distance formula. Thus, shorter distance implies that the node is relatively closer to mid-point of cluster and has higher chance to be chosen as a CH. Secondly, nodes continuously track their energy levels; thus, the nodes with enough energy resource have higher probability to be elected as CHs. Hence, within each cluster, nodes sustain a local table to keep their neighbors information. Consequently, the node with higher neighbors’ density has higher chances to be declared as CH. After obtaining the local information by each node, the values are summed in a weighted manner to determine the rank for CH election based on Eq. 5.

wi = α ei + β

1 dti

+ γ di

(5)

α , β , γ are the coefficients denoting the impact of residual energy, distance and node density factors respectively. The values of all the coefficient are taken in the range of 0 and 1. Also, the coeffiients α , β , γ are assigned in such a way that α + β + γ = 1. Finally, the optimum nodes in terms of residual energy, distance and density factors are potential contestants to be appointed as CHs within the cluster regions.Next, elected CHs flood their local clusterswith their status information in a controlled way by broadcasting an advertisment message. The advertisement message comprises of ID of appointed CH, its position information and CLUS_ID in which it resides. Member nodes upon receiving the message further forwards it among their neighbors till all the nodes in the cluster are informed about CH status. The advertisement message is ignored by nodes if their cluster ID is different. Based on the number of member nodes within each cluster, elected CHs schedule their TDMA cycle by dividing the transmission time Tinto fixed time slots, which aims to avoid chances of data collision Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE 6

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19 Table 2 Routing table of source node. Residual energy ej

Distance to neighbor di, j

Link estimator ETTi, j

Computed FL

——–

———-

———

———

in data transmission and utilize resources in a well-organized manner.CH distributes the transmission slots to its cluster members and each node switch to sleep mode in scheduled manner for saving its energy. 5.2. Dynamic routes setup To ensure timely and reliable data forwarding among nodes, DEFTR constructs parallel routing chains named as Improved Cluster-based Route Discovery (i-CBRD) for both intra-cluster and inter-clusters communication. In addition, to balance the energy consumption over the entire network field, a multi-facet routing function selects an optimized next-hop neighbor for data forwarding by using residual energy, distance and link quality. The rationale behind this is to only select next-hops that can provide shortest, more energy efficient and reliable routing paths. Consequently, such dynamic routes adjustment reduces re-transmissions and balances data traffic loads among nodes, which leads to improving network lifetime and data delivery performance. Firstly, the route setup component initiates a neighbor discovery phase among member nodes to determine nearly optimal intra-cluster data routing paths. Each node floods route request packet among member nodes. Then, adjacent nodes respond with their residual energy and distance information to the source node. In this manner, each node constructs a local table that comprises of collected information of all its adjacent nodes. Moreover, to determine the link quality among neighbors, each node determines the Estimated Transmission Time (ETT) based on Eq. 6, by exploiting Estimated Transmission Count (ETX), that is used to measure the quality of route among two adjacent nodes using the size of the packet (S) and link capacity (B). As a result, each communication link has specified score according to its calculated ETT value. The specified score is traced in the constructed local table by a source node.

E T Ti = E T Xi ∗ S/B

(6)

When a source node tries to determine its next-hop among neighborsNi for data forwarding, it computes the Forwarder Level (FL) by using the recorded information in its local table. The residual energy, distance and ETT are summed in a weighted manner to determine the score for FL by using Eq. 7.

  1 1 F L = α ∗e j + β ∗ +γ∗ , j ∈ Ni Di j

ET Ti j

(7)

α , β , γ are the coefficients denoting the impact of residual energy, distance and link qualtity factors. The values of all the coefficient are taken in the range of [0,1]. Also, the sum of coefficients α , β , γ are assigned in such a way that it must be equal to 1.Thus, each source node selects anappropriate neighbor with the highest FL as its next-hop by exploiting stored information in routing table, as illustrated in Table 2. Afterwards, the source node unicasts Route Request (RREQ) message towards chosen next-hop. The elected next-hop initially confirms whether the destination address is among its neighbors. In the case of finding a match, selected existing node sets the next-hop flag directly towards the destination. If not, the current node selects a new suitable next-hop based on computed FL score and propagates RREQ message towards it. This practice continues till an optimized routing path is constructed towards the corresponding CHs. Furthermore, to decrease excessive energy consumption among CHs, multi-hop based hierarchical formation for dynamic routes are adjusted. The steps undertaken during inter-cluster communication are described as follows. 1. On receiving the BS route discovery information, CHs that are the one-hop distance from BS send their ID’s toward BS with their default transmission power. Consequently, CHs with one-hop distance from BS are capable of direct transmission towards BS thereby constructing tier-1 of the hierarchical structure. 2. In order to keep on the formation of CHs hierarchical structure, CHs at tier-1 further disseminate the information about BS and determine their next-hop CHs. Upon reception of BS message from tier-1, next-hop CHs form the tier-2, send their ID’s towards BS via tier-1 CHs. This same practice is repeated till all undiscovered CHs become a part of the particular tierof hierarchical structure as shown in Fig. 2 3. It might happen that CHs receive multiple propogated messages with different hop counts, in such case; the minimum hop-count is elected as an upper tier for data routing. By this manner, multi-hop routes towards BS are constructed that are utilized for consequent data transmission. 4. On receiving the data packets, if BS is positioned at the next-hop of source CH then aggregated sensory data will be transmitted directly to BS. Otherwise, CH at a lower tier exploits its upper tier CH for data forwarding. In addition, lower tier CHs are not allowed to directly data forwarding towards BS.In this manner, every downstream CH consumes less transmission power while delivering data. 5. It might be a case that there are multiple CHs in the upper tier, candidate CHs is identified by comparing their residual energy to a certain threshold. Afterwards, the ETT value of candidate CHs is computed, and finally, CH with the lowest ETT value is elected as next-hop. Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

[m3Gsc;November 4, 2016;19:8] 7

Fig. 2. Linkage of CHs in different tiers.

S

RREQ A

Status< threshold

RREQ

Primary route

Status
Status< threshold

Alternative route E

RREQ

RREQ

Status< threshold

C

H

G

RREQ

RREQ

F

D

B RREQ

RREQ

discard

I RREQ

RREQ

Fig. 3. Paradigm of intra-cluster disjoint route discovery.

5.3. Fault tolerability Sensor nodes have restricted energy constraints and are extremely prone to faults due to performing numerous functions. The malfunction of nodes not only influences the data forwarding but also reduces network connectivity and lifetime. In our DEFTR protocol, the fault tolerability component is carried out in two phases. The first phase initiates a mechanism to discover an optimized alternative disjoint routing path when encountering low energy nodes on an active route inside each cluster region. The constructed alternative routing path can be used in case of primary route failure. The second phase commences a mechanism for balancing load distribution and energy consumption among elected CHs. Thus, the function of CH is rotated among member nodes by exploiting network conditions. In the first phase, the source node switches to the alternative routing path upon receiving aroute-error message from any of its primary route nodes. The route-error message is sent towards the source when the energy status of an intermediate node drops to predefine threshold. By receiving route-error message, the neighbors towards the source node propagate their information and update routing tables. In DEFTR, the alternative routing paths towards associated CHs are constructed by source nodes in advance, based on FL values for their neighbors, by exploiting the aforementioned dynamic routes setup component. The construction of alternative routes in advance contributes in decreasing route recovery time and data latency. In order to balance the load distribution among intermediate nodes, DEFTR exploits the saved contents of routing table and re-evaluates the FL values. Accordingly, based on new FL values, a source node sends RREQ message to its next most preferred neighbor with different Route ID. Nevertheless, to attain energy efficiency and longer route lifetime, the primary and alternative routes are constructed as disjoint routes by utilizing different intermediate nodes. To achieve this, each node can accept only one RREQ message at a time. In the case of receiving another RREQ while being part of the primary route, the incoming RREQ packet is ignored as shown in Fig. 3, if node H that is part of the primary route, received another route request via node F then RREQ packet is discarded. However, if RREP message is not received at node F within a certain time period (t), node F unicasts the RREQ message to the second most preferred neighbor node based on computed FL value. This procedure is applied repeatedly by intermediary nodes until the route is constructed towards the destination. After switching to the alternative route, the source node sends a route-update message to the primary route nodes in order to terminate the primary route. Afterwards, a source node will consider the alternative route as a new primary route and construct a new alternative route. Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

JID: CAEE 8

ARTICLE IN PRESS

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19 Table 3 Simulation parameters. Parameter

Value

Sensor field Eelect Eamp Efs Packet size Control packet size Initial energy of node Probability of clusters (p) Simulation time Round period Node’s transmission range

20 0 × 20 0m2 50nJ/ bit 10nJ/bit/m2 0.0013pJ/bit/m4 512 bits 25 bits 5J 5% 1500 sec 20 sec 30m

In the second phase, CHs that are nearer to BS suffer higher traffic load, which consumes their energy sooner as compared to farther CHs. As a result, t is calculated based on Eq. 8, whereas t for CHs nearer to BS has a lower value which results in rotating CHs more frequently to balance their energy utilization.

t = t ∗ distchi toBS /distmax

(8)

wheret is fixed round step, distchi toBS is the distance between CH towards BS and distmax is the distance of the farthest node from BS.Furthermore, to increase the energy conservation and fault tolerance of inter-cluster routing, CH energy consumption is exploited as a vital metric in CH rotation. However, DEFTR protocol gives higher precedence to CH energy consumption as compared to its distance from BS. When a CH residual energy falls under a certain threshold (i.e. 50% of energy status when it was elected as CH), a re-election routine begin among its member nodes to choose a new CH. Otherwise, upon reaching the end of t, the chosen CH alters its next-hop flag to false and quit from data forwarding. However, CH reelection process is initiated within specified zone in the cluster region. Current CH computes a threshold distance Dthreshold by exploiting its own distance from the centre of the cluster and floods the region bounded by Dthreshold requesting the nodes to respond with their residual energy and distance. Nodes that are positioned inside the threshold distance, respond to CH with their residual energy and distance. In addition, nodes that are located within the Dthreshold further forward the CH’s request with their neighbors till it is discarded by nodes outside the threshold distance. If current CH does not receive any response from its member nodes within the predefined time interval (t), it assumes that no qualified nodes are found within the computed Dthreshold and accordingly gradually increases the Dthreshold to expand the search zone. In the re-election mechanism, the residual energy of qualified nodes is given high priority as compared to their distance from the centroid of the cluster. Among candidates, the nodes with residual energy greater than the threshold are identified and then the node relatively near to centroid is elected as a new CH. Next, the new chosen CH floods its cluster by sending ID and position information in order to update the transmission schedule and members association. In addition, the new chosen CH informs adjacent CHs by sending its status information for the function of dynamic inter-cluster routing. Accordingly, DEFTR protocol localized the CH rotating mechanism within the cluster region that greatly decreases re-election overheads and improves network connectivity. The three components of DEFTR protocol are governed by Algorithm 1. 6. Simulation and performance analysis This section presents the simulation setup and performance evaluation of our proposed DEFTR protocol. To evaluate the efficacy of DEFTR protocol, a sensitivity analysis is carried out to determine the impact of coefficient factors with different possible assignments. Then, based on varying number of nodes and packets generation rate, the DEFTR protocol is assessed against the existence schemes (MR-LEACH, DFCR, and CCBRP). Table 3 illustrates the simulation parameters that are used in this study and have been exploited by several researchers [21–23] to evaluate the performance of their proposed solutions. We run the simulation for 1500 sec and time interval for each round is set to 20 sec. The performance of proposed DEFTR protocol is evaluated in terms of network lifetime and throughput, transmission cost, and average end-to-end delay. 6.1. Sensitivity of coefficient factors Basically, no optimum values can be given to coefficient factors that would be suitable for all kind of network applications or scenarios. In order to perform sensitivy analysis of CH election mechanism, three different configurations are made such that configuration-1 corresponds to α = 0.6, β = 0.2 and γ = 0.2, configuration-2 denotes α = 0.2, β = 0.6, and γ = 0.2, whereas configuration-3 represents α = 0.2, β = 0.2, and γ = 0.6. Configuration-1, configuration-2, and configuration-3 are evaluated in terms of average energy level, average distance measurement and neighborhood density as a criterion to determine the optimum values for α , β , γ . Fig. 4 shows that in configuration-1 the most energy efficient nodes are elected as CHs where higher weight is given to energy factor α . Fig. 5 depicts the configuration-2 effects where the node closest to the centroid is selected as CH. Fig. 6 describes configuration-3 whereas the node with the highest neighborhood density is selected as CH. Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

JID: CAEE

ARTICLE IN PRESS

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

9

Algorithm 1 Dynamic energy-aware fault tolerant routing protocol. 1st Component: Clusters formation 1. Exploiting network size n and required number of clusters to compute network portioning 2. procedureGENERATION OF CLUSTERS(K) 3. for each partition i ∈ [1:K] 4. do 5. Determine central point CP 6. LCP [i] = CP ; 7. end for 8. for each node i ∈ [1:N] 9. do 10. Computedistance(i x,y , LCP )& joins to 11. nearest centroid 12. end for 13. for each member_nodei,j ∈ [1: clusterj_members ] 14. do 15. wi = α ei + β ( d1t ) + γ di i 16. end for 17. end procedure 2nd Component: Dynamic Route Setup 1. procedure Intra-clusters routes discovery 2. for each node m ∈ [1:M] 3. do mdeterminesFL of adjacent nodes 4. end for 5. while (y!= destination) yselectsRREQ to adjacent node with highest FL 6. end while 7. end procedure 8. procedure Inter-clusters routes discovery 9. for each node i on receiving information of BS 10. do 11. If (Id = = identity of CH) &&next-hop = = BS) 12. store the BS information &forward response 13. else if (Id = = CH id) &&next-hop != BS_ID) 14. nodestore BS information ID& electedas C Hupper_tier 15. else if (node Id! = CH id) 16. ignore information of BS 17. end if 18. end procedure 3rd Component: Fault Tolerance 1. procedure Intra-Cluster fault tolerance routing 2. Ralternative = Intra-Cluster alternative routes construction () 3. if (m Rp and mres < threshold) 4. Send route_err message to ms 5. ms switch to Ralternative 6. end if 7. procedure Intra-Cluster alternative routes construction 8. if (the primary route is discovered) 9. y= ms 10. while (y!= destination) 11. yselects np which has secondmaximum FS value 12. ycreates RREQ and send to np 13. ifnp is part of active RouteID then 14. discard RREQ 15. Repeat step no. 11&12 16. else 17. np Reply to y 18. y = np 19. end if 20. end procedure 21. procedure Inter-clusters fault tolerance 22. if (Chi . energy < threshold) 23. shift the position of (Chi ) with appropriate node 24. else 25. t= t ∗ distchi toBS /distmax 26. shift the position of (Chi ) with appropriate node 27. end if 28. end procedure

Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE 10

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

2 Configuration-1 Configuration-2 Configuration-3

1.9

Average energy level (j)

1.8 1.7 1.6 1.5 1.4 1.3 1.2 2

3

4

5 6 7 Number of cluster Heads

8

9

10

Fig. 4. Average energy v/s cluster heads.

40

Average distance measurement

35

30

25

20

15 Configuration-1 Configuration-2 Configuration-3

10

5

2

3

4

5 6 7 Number of cluster Heads

8

9

10

Fig. 5. Average distance measurement v/s cluster heads.

Consequently, to expose more balanced contribution and optimize the CH election mechanism, the coefficient factors are set to uniform values (α = β = γ = 0.33, where as α + β + γ = 1). In addition, this section presents the analysis of different coefficient factors that are incorporated in routing decisions. In order to perform the sensitivity analysis, three different configurations are made such that i-CBRD-1 corresponds to α = 0.5, β = 0.3 and γ = 0.2, i-CBRD-2 denotes α = 0.2, β = 0.5 and γ = 0.3, whereas i-CBRD-3 represents α = 0.3, β = 0.2 and γ = 0.5. i-CBRD-1, i-CBRD-2 and i-CBRD-3 are evaluated in terms of network lifetime, delivery latency and message delivery ratio to determine the optimum values for α , β , γ . Fig. 7 Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

11

11 Configuration-1 Configuration-2 Configuration-3

10

Average neighborhood density

9 8 7 6 5 4 3 2 1 0

2

3

4

5 6 7 Number of cluster Heads

8

9

10

Fig. 6. Average neighborhood density v/s cluster heads.

500 i-CBRD-1 i-CBRD-2 i-CBRD-3

450

Netwowk lifetime (sec)

400 350 300 250 200 150 100 20

30

40

50 60 70 Number of nodes

80

90

100

Fig. 7. Network lifetime v/s number of nodes.

shows that i-CBRD-1 acheives better network lifetime as high weight is assigned to the energy factor α . Fig. 8 depicts the delivery latency results, where i-CBRD-2 has the best performance due to assigning high weights to the distance factor β . Fig. 9 describes i-CBRD-3, whereas the nodes with highest link quality factor γ are selected as data forwarders. In this research paper, we considered the nodes as faulty or exhausted due to depletion of their energy resources. The optimum selection of intermediate nodes in proposed protocol not only improves data delivery performance in an energy efficient manner but on the other hand, also increases fault tolerability level. To estimate the impact of co-efficient factorsα , β , γ in faulty scenarios, i-CBRD-1, i-CBRD-2 and i-CBRD-3 are evaluated in terms of route lifetime, a number of route discoveries and transmission energy consumption. Fig. 10 shows that i-CBRD-1 acheives better performance in term of route lifetime, beacuse high weight is assigned to energy factor α , which results in electing only the most energy efficient nodes as data forwarders thereby improving route lifetime. Fig. 11 depicts the results for consumed transmission energy during data forwarding, where i-CBRD-2 demonstrates better performance than i-CBRD-1 and i-CBRD-3 due to assigning ahigh Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE 12

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

260 240 220

Data latency (sec)

200 180 160 140

i-CBRD-1 i-CBRD-2 i-CBRD-3

120 100 80 60 20

30

40

50 60 70 Number of nodes

80

90

100

Fig. 8. Delivery latency v/s number of nodes.

100 i-CBRD-1 i-CBRD-2 i-CBRD-3

Message delivery ratio (%)

90

80

70

60

50

40

30 20

30

40

50 60 70 Number of nodes

80

90

100

Fig. 9. Message delivery ratio v/s number of nodes.

weight to distance factor β . The significance of incorporating distance factor during routing decision is to elect the least number of data forwarder, which results in shortening the routing path. Fig. 12 presents the result of routes discovery upon encountering exhausted nodes. i-CBRD-3 illustrates betterperformance than i-CBRD-2 and i-CBRD-1 whereas the nodes with highest link quality factor γ are selected as data forwarders. The integration of link quality factor in routing decision significantly improves data reliability in the faulty scenario and decreases the number of re-transmissions. 6.2. Results analysis under varying network sizes Fig. 13 depicts that our DEFTR protocol exceeds the existing schemes by 9.5%, 13%, and 20.9% respectively in terms of network lifetime. This is due to DEFTR protocol generates clusters on the basis of network size and initiates the CH selection Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

13

100 i-CBRD-1 i-CBRD-2 i-CBRD-3

90

Route lifetime(sec)

80 70 60 50 40 30 20 20

30

40

50 60 70 Number of nodes

80

90

100

Fig. 10. Route lifetime v/s number of nodes.

Average transmission energy consumption(j)

0.22

0.2

0.18

0.16

0.14

i-CBRD-1 i-CBRD-2 i-CBRD-3

0.12

0.1 20

30

40

50 60 70 Number of nodes

80

90

100

Fig. 11. Energy consumption v/s number of nodes.

process inside the bounded regions thereby gradually decreases energy consumption. In addition, DEFTR protocol shifts the role of CHs by exploiting network conditions. It is observed in Fig. 14 that throughput produced by DEFTR protocol is 15.7%, 24.5%, and 35% respectively higher than existing schemes. Existing routing schemes construct the shortest path for data forwarding by using least number of hops. Consequently, highly reduce the achievable network throughput due to low flexibility against node failure, which results in lower overall network performance. Fig. 15 illustrates that DEFTR protocol achieved lower end-to-end delay by 14%, 22.5%, and 29% respectively than existing schemes. The existing solutions exhibit a higher end-to-end delay in data forwarding process because of constructing routing path without considering link quality, which results in a higher route failure probability. Transmission cost is measured over the constructed routing path based Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE 14

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

45

Number of route discovery packets

40 35 30 25 20 15 i-CBRD-1 i-CBRD-2 i-CBRD-3

10 5 20

30

40

50 60 70 Number of nodes

80

90

100

Fig. 12. Route discoveriesv/s number of nodes.

1000 DEFTR MR-LEACH DFCR CCBRP

950

Network lifetime(sec)

900

850

800

750

700

650 100

150

200

250 300 350 Number of nodes

400

450

500

Fig. 13. Impact of number of nodes on network lifetime.

on Eq. 9, while transmitting the data packets between two consecutive nodes.

K=n Ea =

k=1

( E ik − E f k ) n

(9)

Ei denotes node’s initial energy, Ej is node’s residual energy status and n is the number of nodes that are distributed over the network field. It is seen in Fig. 16 that DEFTR protocol reduced the transmission cost by 17.8%, 35.8%, and 46% than existing schemes. In particular, DEFTR protocol transmits data by utilizing energy efficient multi-hop routes, which provides load balancing and reducing communication distance among routes. Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

15

90 DEFTR MR-LEACH DFCR CCBRP

85

Network throughout(kbits/sec)

80 75 70 65 60 55 50 45 40 100

150

200

250 300 350 Number of nodes

400

450

500

Fig. 14. Impact of number of nodes on network throughput.

0.045

Average end-to-end delay(ms)

0.04

0.035

0.03

0.025

DEFTR MR-LEACH DFCR CCBRP

0.02

0.015 100

150

200

250 300 350 Number of nodes

400

450

500

Fig. 15. Impact of number of nodes on end-to-end delay.

6.3. Result analysis under varying packets generation rate Obviously, using higher packet generation rate increases network traffic load, which results in reducing network lifetime. However, as illustrated in Fig. 17, DEFTR protocol improved network lifetime by 11.2%, 18.6%, and 27.1% as compared to existing schemes. This is due to the generation of different clusters based on network size rather than the probabilistic method, and the utilization of composite routing function optimized next-hop selection for propagating sensory information. In addition, it is seen in Fig. 18 that DEFTR improved network throughout by 14%, 22.2%, and 32.6% as compared to existing schemes. This is due to the construction of parallel routing paths and incorporatration of fault tolerability at multi-level, which decreases network disruption and improved network throughput. In Fig. 19, DEFTR decreases the end-to-end delay Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE 16

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

0.5 0.45 0.4

Transmission cost(j)

0.35 0.3 0.25 0.2 0.15 DEFTR MR-LEACH DFCR CCBRP

0.1 0.05 0 100

150

200

250 300 350 Number of nodes

400

450

500

Fig. 16. Impact of number of nodes on transmission cost.

1150 DEFTR MR-LEACH DFCR CCBRP

1100 1050

Network lifetime

1000 950 900 850 800 750 700 650 600 2

3

4

5 6 7 Data packets per second

8

9

10

Fig. 17. Impact varying generation of packets on network lifetime.

rate by 18.5%, 29.7%, and 37%, respectively. This is because of the construction of more reliable routes in terms of energy level and link quality and the construction of alternative routing paths in advance, which reduces the number of re-transmissions and route breakages. It is observed in Fig. 20 that DEFTR protocol reduces the transmission cost during routing by 27%, 33%, and 38.8% as compared to existing schemes. In particular, DEFTR reduces the communication distance between two consecutive nodes, which reduces the required transmission power and energy. 7. Conclusion This paper has presented a Dynamic Energy-aware and Fault Tolerant Routing protocol called DEFTR for WSNs, which aims to improve network lifetime and routing performance. DEFTR partitioned the entire network field into balance-sized Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

17

90 DEFTR MR-LEACH DFCR CCBRP

85

Network throughput(kbits/sec)

80 75 70 65 60 55 50 45 40

2

3

4

5 6 7 Data packets per second

8

9

10

Fig. 18. Impact of varying generation of packets network throughput.

A verage end-to-end delay(ms)

0.04

0.035

0.03

0.025

0.02 DEFTR MR-LEACH DFCR CCBRP

0.015

0.01

2

3

4

5 6 7 Data packets per second

8

9

10

Fig. 19. Impact of varying generation of packets on network delay.

clusters based on network size with minimum energy consumption. In addition, using compound metrics, DEFTR performs localized CHs election mechanism to reduce computational overheads and initiates re-election mechanism on demand basis. By exploiting multi-facet attributes in routing decision, our proposed DEFTR protocol offers a more energy efficient and cost effective solution to deliver data in a reliable way while considering WSN constraints. The proposed protocol is highly tolerant to route instability and excessive network load scenarios. Moreover, providing fault tolerance support for both intracluster and inter-clusters communication significantly improved data network connectivity and decreased data disruption. The simulation results reveal improved performance of DEFTR protocol compared to relevant energy efficient clustering solutions. Our DEFTR protocol presents a lightweight solution for resource constraint WSNs. However, its efficiency on real hardware platform needs to be studied. For future work, a test-bed is going to be developed for more real evaluation of the Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

ARTICLE IN PRESS

JID: CAEE 18

[m3Gsc;November 4, 2016;19:8]

K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

0.28 0.26

Transmission cost (j)

0.24 0.22 0.2 0.18 0.16 DEFTR MR-LEACH DFCR CCBRP

0.14 0.12 0.1

2

3

4

5 6 7 Data packets per second

8

9

10

Fig. 20. Impact of varying generation of packets on transmission cost.

DEFTR protocol. In addition, the proposed protocol can be further improved by exploiting mobility issues of sensor nodes during construction and maintenance of routing paths. Acknowledgement We gratefully acknowledge the technical support and research facilities provided by the Universiti Teknologi Malaysia (UTM), without which this work could not have been completed. References [1] Potdar V, Sharif A, Chang E. Wireless sensor networks: a survey. In: Advanced information networking and applications workshops, 2009. WAINA’09. international conference Bradford; 2009. p. 636–41. [2] Karim L, Nasser N, Sheltami T. A fault-tolerant energy-efficient clustering protocol of a wireless sensor network. Wireless Commun Mob Comput 2014;14(2):175–85. [3] Xu Z, et al. Balancing energy consumption with hybrid clustering and routing strategy in wireless sensor networks. Sensors 2015;15(10):26583–605. [4] Samanta M, Banerjee I. Optimal load distribution of cluster head in fault-tolerant wireless sensor network. in Electrical. In: Electronics and computer science (SCEECS), 2014 IEEE students’ conference. Bhopal; 2014. p. 1–7. [5] Latif K, et al. Divide-and-rule scheme for energy efficient routing in wireless sensor networks. Proc Comput Sci 2013;19:340–7. [6] Liu J-L, Ravishankar CV. LEACH-GA: genetic algorithm-based energy-efficient adaptive clustering protocol for wireless sensor networks. Int J Mach Learn Comput 2011;1(1):79–85. [7] Baranidharan B, Srividhya S, Santhi B. Energy efficient hierarchical unequal clustering in wireless sensor networks. Indian J Sci Technol 2014;7(3):301. [8] Selvi GV, Manoharan R. Balanced unequal clustering algorithm for wireless sensor network. i-Manager’s J Wireless Commun Netw 2015;3(4):327–32. [9] Issariyakul T, Hossain E. An introduction to network simulator NS2. Springer; 2012. [10] Heinzelman WR, Chandrakasan A, Balakrishnan H. Energy-efficient communication protocol for wireless microsensor networks. In: System Sciences, 20 0 0. Proceedings of the 33rd annual hawaii international conference. Maui; 20 0 0. p. 1–10. [11] Jannatul Ferdous M, Ferdous J, Dey T. Central base-station controlled density aware clustering protocol for wireless sensor networks. In: Computers and information technology, 2009. ICCIT’09. 12th international conference. Dhaka; 2009. p. 37–43. [12] Xinhua W, Sheng W. Performance comparison of LEACH and LEACH-C protocols by NS2. In: Distributed computing and applications to business engineering and science (DCABES), 2010 Ninth international symposium. Hong Kong; 2010. p. 254–8. [13] Farooq MO, Dogar AB, Shah GA. MR-LEACH: multi-hop routing with low energy adaptive clustering hierarchy. In: Sensor technologies and applications (SENSORCOMM), 2010 Fourth international conference. Venice; 2010. p. 262–8. [14] Ali SA, Refaay SK. Chain-chain based routing protocol. IJCSI Int J Comput Sci Issues 2011;8(3):105–12. [15] Chen G, et al. An unequal cluster-based routing protocol in wireless sensor networks. Wireless Netw 2009;15(2):193–207. [16] Ever E, et al.. In: UHEED-an unequal clustering algorithm for wireless sensor networks; 2012. p. 1–9. [17] Alim MA, Wu Y, Wang W. A fuzzy based clustering protocol for energy-efficient wireless sensor networks. ICCSEE 2013;760–762:685–90. [18] Ortiz AM, et al. Fuzzy-logic based routing for dense wireless sensor networks. Telecommun Syst 2013;52(4):2687–97. [19] Kim KT, et al. Tree-based clustering (TBC) for energy efficient wireless sensor networks. In: Advanced information networking and applications workshops (Waina), 2010 IEEE 24th international conference. Yichang; 2010. p. 680–5. [20] Azharuddin M, Kuila P, Jana PK. Energy efficient fault tolerant clustering and routing algorithms for wireless sensor networks. Comput Elect Eng 2015;41:177–90. [21] Gu X, et al. ECDC: an energy and coverage-aware distributed clustering protocol for wireless sensor networks. Comput Elect Eng 2014;40(2):384–98. [22] Mahajan S, Malhotra J, Sharma S. An energy balanced QoS based cluster head selection strategy for WSN. Egypt Inf J 2014;15(3):189–99. [23] Wang S-S, Chen Z-P. LCM: a link-aware clustering mechanism for energy-efficient routing in wireless sensor networks. Sens J IEEE 2013;13(2):728–36.

Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017

JID: CAEE

ARTICLE IN PRESS K. Haseeb et al. / Computers and Electrical Engineering 000 (2016) 1–19

[m3Gsc;November 4, 2016;19:8] 19

Khalid Haseeb received his MS-IT degree from Institute of Management Sciences, Pakistan. He did his Ph.D in Computer Science from Faculty of Computing at Universiti Teknologi Malaysia (UTM), in 2016. Since 2008, he is engaged as a Lecturer in Department of Computer Science, Islamia College Peshawar, Pakistan. His research areas include network security, cloud computing, ad-hoc and wireless sensor networks. Kamalrulnizam Abu Bakar is a Professor and Deputy Dean (Development & Innovation) in Faculty of Computing at Universiti Teknologi Malaysia. He did his PhD in Computer Science from Aston University, United Kingdom, in 2004. He involves in several research projects and is the reviewer for many scientific journals. His specialization includes mobile computing, information security, ad-hoc and wireless sensor networks. Abdul Hanan Abdullah is a Professor in Computer Science at Universiti Teknologi Malaysia. He did his PhD degree from Aston University in Birmingham in 1995. He is involved in several research projects and is the referee for many scientific journals and conferences. His specialization includes grid computing, wireless sensor networks, ad hoc and sensor networks. Adnan Ahmed is an Assistant Professor in the department of Computer Systems Engineering at Quaid-e-Awam University of Engineering, Science and Technology, Nawabshah, Pakistan. He completed his PhD in Computer Science from Universiti Teknologi Malaysia (UTM) in November 2015. His research interest includes ad hoc networks, security and trust management, quality of service and routing issues in wireless body area network. Tasneem Darwish is a Ph.D. student at the Faculty of Computing, Universiti Teknologi Malaysia. She is a member of the Pervasive Computing research group. She received her M.Sc. degree in electronics and electrical engineering from the University of Glasgow, United Kingdom, in 2007. Her current research focuses on vehicular communications, wireless ad hoc networks, and mobile computing. Fasee Ullah graduated from Department of Computer Science, Islamia College Peshawar, Pakistan in March 2007. He has done MS in Computer Networks and Information Technology from SZABIST, Islamabad, Pakistan in 2009. Currently, he is pursuing his PhD at Faculty of Computing, Universiti Teknologi Malaysia (UTM), Malaysia. His research areas include Wireless Body Area Network, sensor and ad-hoc networks and nanotechnology.

Please cite this article as: K. Haseeb et al., A dynamic Energy-aware fault tolerant routing protocol for wireless sensor networks✰, Computers and Electrical Engineering (2016), http://dx.doi.org/10.1016/j.compeleceng.2016.10.017