This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2018.2841911, IEEE
Communications Letters

Traffic and Energy Aware Routing for

Heterogeneous Wireless Sensor Networks
Deepak Sharma, Member, IEEE, and Amol P Bhondekar Member, IEEE

role.
Abstract— The energy-efficiency of routing algorithm is The heterogeneity in terms of disparities in data generation
crucial for improving the lifetime of battery constrained Wireless rate (traffic) is considered under computation heterogeneity
Sensor Networks (WSNs). The consideration of nodes [5]. Sharma et al. [6] analyzed the effect of traffic
heterogeneity in routing is essential for achieving optimal
heterogeneity in homogeneous WSN routing (LEACH)
resource utilization. This letter considers sensor nodes with
random initial energies and random disparities in data algorithm. Energy Dissipation Forecast and Clustering
generation rate (traffic) to model a realistic clustering based Management (EDFCM) [5] considers traffic heterogeneity
WSN suited for heterogeneous sensing applications. The letter along with energy heterogeneity in a very specific two-level
presents an energy model for the scenario and proposes a Traffic WSN. Further, EDFCM considers additional nodes
and Energy Aware Routing (TEAR) scheme to improve the (management nodes) to control the number of clusters, which
stability period. The simulation results indicate that TEAR
makes its natural distributed localized decision-making
outperforms other clustering based routing algorithms under the
scenario. behavior questionable. The consideration of traffic
heterogeneity along with energy heterogeneity is crucial for
Index Terms— Wireless sensor networks, routing protocols, modeling realistic WSNs with application heterogeneity and
heterogeneity, clustering, stability period event-driven scenarios.
This letter considers both, energy and traffic
I. INTRODUCTION heterogeneities, with multiple random levels. An energy
model is presented for the multi-heterogeneity scenario, where
I nternet of Things (IoT) envisions interoperability of
heterogeneous devices to support diverse applications, and
the Wireless Sensor Network (WSN) technology is an
consideration of multi-level traffic heterogeneity is a novel
concept. A novel routing algorithm named Traffic and Energy
important building block of IoT sphere. Consideration of Aware Routing (TEAR) is presented, which considers node’s
heterogeneity (e.g., energy, link and computational traffic requirements along with its energy levels while making
heterogeneities) [1] can improve the performance of WSN CH selection. TEAR shows improvements in terms of stability
routing algorithms in terms of network lifetime, stability, period (reliable lifespan of the WSN before the death of its
reliability, network delay, etc. The energy heterogeneity in first node) over existing algorithms (LEACH, SEP and DEEC)
WSN routing is pursued widely; however, the link and under the scenario.
computation heterogeneities, which are generally used along The rest of this letter is arranged as follows. Section II
with the energy heterogeneity, are relatively less explored presents the system model, which includes the energy model
areas. for the multi-heterogeneous scenario. In Section III, the
In the early work in WSN routing algorithms for energy proposed routing algorithm is described. The simulation
heterogeneous scenarios, Stable Election Protocol (SEP) [2] results have been discussed in Section IV. Finally, Section V
considers two-level energy heterogeneity in Low-Energy concludes the letter.
Adaptive Clustering Hierarchy (LEACH) [3] like cluster-head
(CH) role rotation environment. SEP proposes weighted II. SYSTEM MODEL
election probabilities based on the initial energies of the nodes Considering the basic radio model [3], the transmitter (𝑇𝑥)
to give energy-rich nodes more chances of becoming CHs. considers energy dissipation in the radio electronics and the
The Distributed Energy-Efficient Clustering (DEEC) [4] power amplifier, and the receiver (𝑅𝑥) considers the radio
considers multi-level energy heterogeneous WSN and prefers electronics dissipation. The energy spent in transmitting an
nodes with higher initial energy and residual energy for CH 𝑚𝑖 -bits message over a distance 𝑑 is given by
𝑚𝑖 . 𝐸𝑒𝑙𝑒 + 𝑚𝑖 . 𝜖𝑓𝑠 . 𝑑 2 𝑖𝑓 𝑑 < 𝑑0 (1)
𝐸𝑇𝑥 (𝑚𝑖 , 𝑑) = �
Manuscript received February 15, 2018; revised April 15, 2018; accepted 𝑚𝑖 . 𝐸𝑒𝑙𝑒 + 𝑚𝑖 . 𝜖𝑚𝑝 . 𝑑 4 𝑖𝑓 𝑑 ≥ 𝑑0
May 15, 2018. Where 𝐸𝑒𝑙𝑒 is the electronic circuit’s per bit energy
The authors are with the CSIR-Central Scientific Instruments Organisation,
Chandigarh, India and the Academy of Scientific and Innovative Research, dissipation of the transmitter or the receiver, and the per bit
CSIR-CSIO campus, India (e-mail: [email protected]). energy dissipation in transmitter amplifier is represented by
𝜖𝑓𝑠 . 𝑑 2 or 𝜖𝑚𝑝 . 𝑑 4 depending on the free space or the multipath
transmitter amplifier model respectively (based on 𝑑0 =

1089-7798 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2018.2841911, IEEE
Communications Letters

𝜖𝑓𝑠 𝑘. 𝐸𝐶𝐻 = ∑𝑁−𝑘 𝑁

𝑖=1 ( 𝑚𝑖 . 𝐸𝑒𝑙𝑒 ) + ∑𝑖=1( 𝑚𝑖 . 𝐸𝐷𝐴 ) +
�𝜖 ). 𝜖𝑓𝑠 and 𝜖𝑚𝑝 are fixed radio parameter related to 𝑇𝑥 𝑁 2
(9)
𝑚𝑝 ∑𝑖=𝑁−𝑘�𝑚𝑚𝑎𝑥 . 𝐸𝑒𝑙𝑒 + 𝑚𝑚𝑎𝑥 . 𝜖𝑓𝑠 . 𝑑𝑡𝑜𝐵𝑆 �
amplifier's energy dissipation in free space and multipath Where 𝐸𝐷𝐴 is per bit energy spent in data aggregation.
scenarios respectively. The energy dissipated in receiving an 𝑁−𝑘 𝑁 𝑁−𝑘
Considering ∑𝑁−𝑘
𝑖=1 𝛼𝑡ℎ 𝑖 ≈ ∑𝑖=1 𝛼𝑡ℎ 𝑖 ≈ 𝛼 𝑇𝑜𝑡 , then
𝑚𝑖 -bits message is given by 𝑁 𝑁
𝑁−𝑘
𝐸𝑅𝑥 (𝑚𝑖 ) = 𝑚𝑖 . 𝐸𝑒𝑙𝑒 (2) 𝑘. 𝐸𝐶𝐻 = 𝑚0 (𝑁 + 𝛼 𝑇𝑜𝑡 )𝐸𝑒𝑙𝑒 + 𝑚0 (𝑁 +
𝑁
(10)
𝛼 𝑇𝑜𝑡 ) 𝐸𝐷𝐴 + 𝑘. 𝑚0 (1 + 𝛼𝑡ℎ )𝐸𝑒𝑙𝑒 + 𝑘. 𝑚0 (1 +
Considering the clustering based approach [2-6], with 𝑁 2
𝛼𝑡ℎ )𝜖𝑓𝑠 . 𝑑𝑡𝑜𝐵𝑆
heterogeneous nodes uniformly distributed (spatially) in a
Based on (6), (8) and (10), the total energy dissipated in one
square region (𝑅 × 𝑅). The base-station/sink (BS) is located at
round is given by
the centre of the region [2, 4, 5] and all the nodes are within a 𝑁−𝑘
range of distance 𝑑0 (free space model) from the BS. The 𝐸𝑅𝑜𝑢𝑛𝑑 = 𝑚0 (𝑁 + 𝛼 𝑇𝑜𝑡 )𝐸𝑒𝑙𝑒 + 𝑚0 (𝑁 +
𝑁
nodes send their data to their respective CH, which aggregates 𝛼 𝑇𝑜𝑡 ) 𝐸𝐷𝐴 + 𝑘. 𝑚0 (1 + 𝛼𝑡ℎ )𝐸𝑒𝑙𝑒 + 𝑘. 𝑚0 (1 + (11)
the member nodes’ data and forwards the aggregated message 2 𝑅2
𝛼𝑡ℎ )𝜖𝑓𝑠 . 𝑑𝑡𝑜𝐵𝑆 + 𝑚0 (𝑁 + 𝛼 𝑇𝑜𝑡 )(𝐸𝑒𝑙𝑒 + 𝜖𝑓𝑠 )
2𝜋𝑘
to the BS.
The optimal number of clusters per round (𝑘𝑜𝑝𝑡 ) can be
The initial energies of the nodes are randomly distributed
calculated by finding the derivative of 𝐸𝑅𝑜𝑢𝑛𝑑 (11) with
over [𝐸0 , 𝐸0 (1 + 𝛼𝑒ℎ )] to represent a realistic energy
respect to 𝑘 and setting it to zero.
heterogeneous deployment, where 𝐸0 is the lower bound and
𝑁(𝑁+𝛼𝑇𝑜𝑡 )𝜖𝑓𝑠 .𝑅 2
𝛼𝑒ℎ is the energy heterogeneity factor controlling the upper 𝑘𝑜𝑝𝑡 = 𝑘 = � 2
(12)
2𝜋( (𝑁.𝛼𝑡ℎ −𝛼𝑇𝑜𝑡 )𝐸𝑒𝑙𝑒 +𝑁(1+𝛼𝑡ℎ )𝜖𝑓𝑠 .𝑑𝑡𝑜𝐵𝑆 )
bound. The total initial energy of the WSN is given by
In the absence of traffic heterogeneity, 𝛼𝑡ℎ = 𝛼 𝑇𝑜𝑡 = 0, then
𝐸𝑇𝑜𝑡 = ∑𝑁 𝑖=1 𝐸0 �1 + 𝛼𝑒ℎ 𝑖 �
(3)
𝑁.𝑅 2 𝑁 𝑅 (13)
Where 𝛼𝑒ℎ 𝑖 is the energy heterogeneity factor for node 𝑖. 𝑘𝑜𝑝𝑡 = � 2 = �
2𝜋(𝑑𝑡𝑜𝐵𝑆 ) 2𝜋 𝑑𝑡𝑜𝐵𝑆
Further, to support heterogeneous sensing/computing Which is the optimum number of clusters for a non traffic
requirements in terms of data-generation/traffic (i.e. the heterogeneous WSN [2]. Further based on (12) and (13), the
number of bits in data-messages), the data-message length of value of 𝑘𝑜𝑝𝑡 is lesser for traffic heterogeneous scenarios.
the node 𝑖 with traffic heterogeneity factor 𝛼𝑡ℎ 𝑖 is given by
𝑚𝑖 = 𝑚0 �1 + 𝛼𝑡ℎ 𝑖 �, which is randomly distributed III. PROPOSED ROUTING ALGORITHM
over [𝑚0 , 𝑚0 (1 + 𝛼𝑡ℎ )], where 𝑚0 is the lower bound and 𝛼𝑡ℎ This section first discusses in brief the effects of energy and
is the traffic heterogeneity factor controlling the upper bound. traffic heterogeneities, which provides insight for an effective
It is assumed that the system fulfils the bandwidth CH selection in multi-heterogeneity scenario. Then, the
requirements to support such heterogeneity. proposed routing protocol is presented, which considers
Based on [2-4], for uniformly distributed nodes the average nodes’ initial energy, residual energy and traffic load along
distance between the cluster member nodes and the CH with the average energy of the round during CH selection.
(𝑑𝑡𝑜𝐶𝐻 ), and the average distance between the CHs and the BS
A. Traffic and Energy Heterogeneities in WSN
(𝑑𝑡𝑜𝐵𝑆 ) are given by
𝑅 An increase in traffic heterogeneity, by increasing nodes’
𝑑𝑡𝑜𝐶𝐻 = (4)
√2𝜋𝑘 packet lengths, increases the effective number of bits per
𝑅
𝑑𝑡𝑜𝐵𝑆 = 0.765 (5) round for communication. This increases the WSN energy
2
Where 𝑘 represents the number of clusters. Based on [3], consumption per round and reduces the WSN lifetime (and the
the total energy dissipated in one round is given by stability period). The effect is discussed further in Section IV
𝑁 based on simulation results.
𝐸𝑅𝑜𝑢𝑛𝑑 = 𝑘 �𝐸𝐶𝐻 + � − 1� 𝐸𝑛𝑜𝑛𝐶𝐻 � (6)
𝑘
≈ 𝑘. 𝐸𝐶𝐻 + 𝑁. 𝐸𝑛𝑜𝑛𝐶𝐻
Where 𝐸𝐶𝐻 and 𝐸𝑛𝑜𝑛𝐶𝐻 are energies dissipated in a CH
node and a non-CH node respectively. The energy dissipated
in 𝑁 non-CH nodes in a round is given by
𝑁
2
𝑁. 𝐸𝑛𝑜𝑛𝐶𝐻 = � ( 𝑚𝑖 . 𝐸𝑒𝑙𝑒 + 𝑚𝑖 . 𝜖𝑓𝑠 . 𝑑𝑡𝑜𝐶𝐻 ) (7)
𝑖=1
2
= 𝑚0 (𝐸𝑒𝑙𝑒 + 𝜖𝑓𝑠 . 𝑑𝑡𝑜𝐶𝐻 )(𝑁 + ∑𝑁𝑖=1 𝛼𝑡ℎ 𝑖 )
Considering (4), (7) and 𝛼 𝑇𝑜𝑡 = ∑𝑁 𝛼
𝑖=1 𝑡ℎ 𝑖
𝑅2 (8)
𝑁. 𝐸𝑛𝑜𝑛𝐶𝐻 = 𝑚0 (𝑁 + 𝛼 𝑇𝑜𝑡 )(𝐸𝑒𝑙𝑒 + 𝜖𝑓𝑠 )
2𝜋𝑘
Considering the CH aggregated message which is sent from
any CH to the BS is 𝑚𝑚𝑎𝑥 = 𝑚0 (1 + 𝛼𝑡ℎ ) bits long. The Fig. 1. Energy consumption pattern of traffic heterogeneous nodes
energy dissipated in the 𝑘 CH nodes in one round in receiving The nodes residual energies are analyzed over the WSN
𝑁 − 𝑘 member nodes data, aggregating the information and lifetime for different traffic heterogeneous scenarios (i.e. for
transmitting 𝑘 CHs data to the BS is given by different 𝛼𝑡ℎ with 𝛼𝑒ℎ = 0). Fig. 1 shows the energy

1089-7798 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2018.2841911, IEEE
Communications Letters

consumption pattern over the rounds of operation for a traffic drain in a uniform manner and few nodes remain alive for
heterogeneous scenario (𝛼𝑡ℎ = 2; 𝛼𝑒ℎ = 0) in DEEC 𝑟 > 𝑅. Based on (16), when 𝑟 approaches 𝑅, 𝐸𝐴𝑣𝑔 (𝑟) becomes
environment. 𝐸𝑖 (𝑟) is the residual energy of node 𝑖 for the a very small quantity and for 𝑟 > 𝑅 it becomes a negative
round 𝑟. It shows that the nodes with higher traffic load (i.e. quantity. In DEEC, 𝑅 is considered 1.5 times of the estimated
higher 𝛼𝑡ℎ𝑖 ) lose their energies faster in comparison to the value to avoid the situation where the last few remaining
nodes with lower traffic loads over the rounds of operation. nodes stay alive and do not form clusters. Many approaches
Under two-level energy heterogeneous WSN, SEP performs have been proposed in the literature to improve the accuracy
better than LEACH by preferring nodes with higher initial of estimated energy per round, e.g., SEARCH [7] considers a
energy for CH role. DEEC performs better than LEACH and semi-centralized approach, where BS keeps track of alive
SEP under multi-level energy heterogeneous WSN by nodes and their residual energies to estimate the average
preferring nodes (for CH role) with higher initial and residual
residual energy of the network over the rounds of operation.
energies over the average energy of the round.
This letter focuses on heterogeneity aspects and a simple
B. Traffic and Energy Aware Routing (TEAR) approach is applied to handle the scenario. The 𝐸𝐴𝑣𝑔 (𝑟) is
The CH selection in TEAR is based on the CH role rotation considered as the value 𝐸𝐴𝑣𝑔 (0.9𝑅) for the rounds 𝑟 > 0.9𝑅 to
approach [2-4], where the node 𝑖 becomes a CH in the current ensure active participation of remaining nodes in cluster
round 𝑟, if the random number selected by the node 𝑖 is less formation for the remaining rounds. This is a better approach
than the threshold T(i, r). for distributed decision-making as nodes are aware of 𝑅 and it
𝑝𝑖 (𝑟) can handle the scenarios, where 𝑟 is much greater than 𝑅. The
1 𝑖𝑓 𝑛𝑜𝑑𝑒 𝑖 ∈ 𝐺(𝑟) (14)
𝑇(𝑖, 𝑟) = �1−𝑝𝑖(𝑟)(𝑟 𝑚𝑜𝑑 𝑝𝑖 (𝑟)
)
values of 𝐸𝑇𝑜𝑡 and 𝑅 are calculated and supplied (through BS
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 broadcast message or node’s initial settings) to the nodes
Where 𝑝𝑖 (𝑟) is the CH selection probability for node 𝑖 during before the beginning of WSN operations.
round 𝑟. 𝐺(𝑟) is a set of eligible nodes for the round 𝑟, where TABLE I
the rotating epoch for node 𝑖 to become eligible again is SIMULATION PARAMETERS
1/𝑝𝑖 (𝑟). DEEC considers randomly distributed energy Parameter Value
heterogeneity and prefers nodes with higher initial and
Number of sensor nodes (𝑁) 100
residual energies for CH role, i.e. an energy-rich node has
WSN Area (𝑅 × 𝑅) 100m x 100m
higher 𝑝𝑖 (𝑟) and higher chances of becoming CH. As the Initial energy lower bound (𝐸0 ) 0.5J
operations of a CH are energy intensive, preferring nodes with Energy consumed in Tx/Rx electronics (𝐸𝑒𝑙𝑒 ) 50 nJ/bit
higher initial energies and higher residual energies improves 𝑇𝑥 Amplifier energy dissipation in free space 10 pJ/bit/m2
scenario (𝜖𝑓𝑠 )
the life of energy weaker nodes and hence it improves the 𝑇𝑥 Amplifier energy dissipation in Multipath 0.0013 pJ/bit/m4
WSN stability period. Section IIIA discusses that an increase scenario (𝜖𝑚𝑝 )
in traffic loads increases the effective number of bits to be Energy consumed in Data Aggregation (𝐸𝐷𝐴 ) 5 nJ/bit/signal
communicated to the BS and hence increases network energy Packet length lower bound (𝑚0 ) 4000 bits
consumption. In traffic heterogeneous scenario, the rate of
energy consumption is higher for the nodes with higher traffic IV. RESULTS AND DISCUSSION
loads. So, it is logical that such nodes should be avoided for The simulation setup considers 100 nodes (𝑁), with
energy intensive operation, e.g., CH role. For a realistic WSN randomness in energy and traffic levels, deployed uniformly in
model, with the nodes having heterogeneous initial energies a 100m x 100m (R × R) area with BS located at the centre of
and data traffic requirements, the proposed algorithm (TEAR) the region. The system model for the multi-heterogeneity
prefers the nodes with higher energies (initial and residual) approach is based on Section II. All the scenarios have been
and avoids the nodes with higher traffic loads for CH role. In simulated in MATLAB and the simulation parameters are
TEAR, the probability of becoming CH for node 𝑖 during detailed in TABLE I. LEACH and SEP have been modified to
round 𝑟 is defined as support multi-level energy heterogeneity based on [4]. Further,
𝑝𝑜𝑝𝑡 .𝑁�1+𝛼𝑒ℎ 𝑖 � 𝑁(1+𝛼𝑡ℎ −𝛼𝑒ℎ 𝑖 )𝐸𝑖 (𝑟)
𝑝 (𝑟) = (15) the algorithms are customized to support energy consumption
𝑖 �𝑁+∑𝑁
𝑖=1 𝛼𝑒ℎ 𝑖 �(𝑁+𝑁𝛼𝑡ℎ −𝛼𝑇𝑜𝑡 )𝐸𝐴𝑣𝑔 (𝑟)
in multi-level traffic heterogeneity, where nodes consider their
Where 𝐸𝐴𝑣𝑔 (𝑟) is average energy of the round and 𝑝𝑜𝑝𝑡 is
specific traffic and the aggregated message sent from CH to
optimal probability of a node to become CH, given by BS is 𝑚𝑚𝑎𝑥 bits long. To handle the traffic heterogeneity in
𝑘
𝑝𝑜𝑝𝑡 = 𝑜𝑝𝑡. The remaining functionality of TEAR is similar DEEC, it has been extended based on the above sections
𝑁
to DEEC. Further, in the absence of traffic heterogeneity, (except the proposed probability function for TEAR).
TEAR falls back to DEEC behaviour. Based on DEEC, the Fig. 2 shows the effect of multi-level traffic heterogeneity
𝐸𝐴𝑣𝑔 (𝑟) is given by on LEACH, SEP and DEEC algorithms in the multi-level
1 𝑟 𝐸 energy heterogeneous scenario. An increase in node packet
𝐸𝐴𝑣𝑔 (𝑟) = 𝐸𝑇𝑜𝑡 �1 − � ; 𝑤ℎ𝑒𝑟𝑒 𝑅 = 𝑇𝑜𝑡 (16)
𝑁 𝑅 𝐸𝑅𝑜𝑢𝑛𝑑 size (from 𝛼𝑡ℎ = 0 to 𝛼𝑡ℎ = 2), while maintaining the energy
Where 𝑅 is the estimated value of network lifetime in terms of heterogeneity (𝛼𝑒ℎ = 3), deteriorates the stable region of all
the number of rounds based on uniform energy drainage in the three algorithms.
each round. In actual scenario, the network energy may not

1089-7798 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2018.2841911, IEEE
Communications Letters

does not perform well with increase in traffic loads. DEEC

better handles (over SEP) the increase in traffic heterogeneity,
as it also considers node’s residual energy. TEAR shows
improved performance with increased traffic as it additionally
considers the traffic loads while making CH decision. In the
absence of traffic heterogeneity, TEAR performs like DEEC.
However, TEAR performs better than all the three algorithms
in the presence of traffic heterogeneity.
TABLE II
STABILITY PERIOD: MEAN (STANDARD DEVIATION)
𝜶𝐞𝒉 , 𝜶𝒕𝒉 LEACH SEP DEEC TEAR
3, 0 1131(56) 1352(63) 1870(99) 1870(99)
2, 2 486(31) 513(51) 658(52) 708(49)
Fig. 2. Effect of traffic heterogeneity on WSN routing algorithms 2, 3 376(27) 386(43) 507(34) 538(40)
1, 2 475(27) 450(36) 579(34) 632(29)
The performance of proposed algorithm (TEAR) under 1, 3 367(23) 339(18) 438(28) 489(27)
multi-heterogeneity scenario is depicted in Fig. 3 for (𝛼𝑡ℎ = 1, 4 295(19) 284(12) 344(31) 392(36)
4; 𝛼𝑒ℎ = 1), where the stability periods are 285, 280, 291 and
379 for LEACH, SEP, DEEC and TEAR respectively. Fig. 4 V. CONCLUSION
shows a similar scenario for (𝛼𝑡ℎ = 2; 𝛼𝑒ℎ = 1), where the Consideration of multi-heterogeneity in WSN routing
stability periods are 521, 448, 567 and 614 for LEACH, SEP, algorithms can help in achieving optimal resource utilization
DEEC and TEAR respectively. TEAR shows improvement in in realistic scenarios. This letter considers WSN nodes with
the stability period over LEACH, SEP and DEEC algorithms random levels of energy and traffic heterogeneities. It devises
under the multi-heterogeneity scenario. The detailed results of a traffic and energy aware routing (TEAR) technique with an
TEAR under different heterogeneity parameter (𝛼𝑡ℎ and 𝛼𝑒ℎ ) improved CH selection method, which considers node’s traffic
are shown in TABLE II, where each instance (mean and along with its initial energy and residual energy. TEAR
standard deviation) is based on analyzing the algorithm on a performs better, in terms of stability period, over legacy
set of ten random WSN deployments. algorithms (LEACH, SEP and DEEC) in the multi-
heterogeneous scenario. Further, the multi-heterogeneity
concept (especially the traffic heterogeneity consideration)
could be helpful in developing more effective routing
algorithms for realistic WSNs and Internet of Things
applications with heterogeneous sensing requirements.

REFERENCES
[1] S. Tanwar, N. Kumar, and J. J. Rodrigues, "A systematic review
on heterogeneous routing protocols for wireless sensor network,"
Journal of network and computer applications, vol. 53, pp. 39-56,
2015.
[2] G. Smaragdakis, I. Matta, and A. Bestavros, "SEP: A stable
election protocol for clustered heterogeneous wireless sensor
Fig. 3. Stability period in Multi-heterogeneous WSN (𝛼𝑡ℎ = 4; 𝛼𝑒ℎ = 1) networks," in Second international workshop on sensor and actor
network protocols and applications (SANPA 2004), 2004.
[3] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan, "An
application-specific protocol architecture for wireless microsensor
networks," Wireless Communications, IEEE Transactions on, vol.
1, pp. 660-670, 2002.
[4] L. Qing, Q. Zhu, and M. Wang, "Design of a distributed energy-
efficient clustering algorithm for heterogeneous wireless sensor
networks," Computer communications, vol. 29, pp. 2230-2237,
2006.
[5] H. Zhou, Y. Wu, Y. Hu, and G. Xie, "A novel stable selection and
reliable transmission protocol for clustered heterogeneous wireless
sensor networks," Computer communications, vol. 33, pp. 1843-
1849, 2010.
[6] D. Sharma, A. P. Bhondekar, A. Ojha, A. Shukla, and C.
Ghanshyam, "A traffic aware cluster head selection mechanism for
hierarchical wireless sensor networks routing," in IEEE Parallel,
Distributed and Grid Computing (PDGC), 2016 Fourth
Fig. 4. Stability period in Multi-heterogeneous WSN (𝛼𝑡ℎ = 2; 𝛼𝑒ℎ = 1) International Conference on, 2016, pp. 673-678.
[7] M.-Y. Wang, J. Ding, W.-P. Chen, and W.-Q. Guan, "SEARCH: A
The results show that the extended version of SEP (based stochastic election approach for heterogeneous wireless sensor
on [4]) do not perform well under traffic heterogeneity networks," Communications Letters, IEEE, vol. 19, pp. 443-446,
scenarios. As SEP considers initial energies in CH selection, it 2015.

1089-7798 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.