CC BY-NC-ND
0Author's Accepted Manuscript
| Digital ...11«. fcl Communications mJ V^? | ^ and Networks
Topology control of tactical wireless sensor networks using energy efficient zone routing
Preetha Thulasiraman, Kevin A. White
www.elsevier.comlocate/dcan
PII: S2352-8648(16)00003-1
DOI: http ://dx.doi. org/ 10.1016/j. dcan.2016.01.002
Reference: DCAN32
To appear in: Digital Communications and Networks
Received date: 12 June 2015 Revised date: 3 December 2015 Accepted date: 12 January 2016
Cite this article as: Preetha Thulasiraman and Kevin A. White, Topology control of tactical wireless sensor networks using energy efficient zone routing, Digita Communications and Networks, http://dx.doi.org/10.1016/j.dcan.2016.01.002
This is a PDF file of an unedited manuscript that has been accepted fo publication. As a service to our customers we are providing this early version o the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form Please note that during the production process errors may be discovered whic could affect the content, and all legal disclaimers that apply to the journal pertain
Topology Control of Tactical Wireless Sensor Networks Using Energy Efficient Zone
Routing
Preetha Thulasiraman, Member, IEEE and Kevin A. White Department of Electrical and Computer Engineering Naval Postgraduate School, Monterey, CA Email: pthulas1@nps.edu, Kevin.A.White@me.navy.mil
Correspondence: Professor Preetha Thulasiraman
Department of Electrical and Computer Engineering
Naval Postgraduate School
833 Dyer Road, Spanagel Hall Rm 448C
Monterey, CA, USA 93940
Tel: (831) 656-3456
Fax: (831) 656-2760
Email: pthulas1@nps.edu
Topology Control of Tactical Wireless Sensor Networks Using Energy Efficient Zone Routing
Preetha Thulasiraman and Kevin A. White
Department of Electrical and Computer Engineering Naval Postgraduate School, Monterey, CA email: pthulas1@nps.edu, Kevin.A.White@me.navy.mil
Abstract
The US Department of Defense (DoD) routinely uses wireless sensor networks (WSN) for military tactical communications. Sensor node die-out has a significant impact on the topology of a tactical WSN. This is problematic for military applications where situational data is critical to tactical decision making. To increase the amount of time all sensor nodes remain active within the network and to control the network topology tactically, energy efficient routing mechanisms must be employed. In this paper, we aim to provide realistic insights on the practical advantages and disadvantages of using established routing techniques for tactical WSNs. We investigate the following established routing algorithms: direct routing, minimum transmission energy (MTE), low energy adaptive cluster head routing (LEACH), and zone clustering. Based on the node die out statistics observed with these algorithms and the topological impact the node die outs have on the network, we develop a novel, energy efficient zone clustering algorithm called EZone. Via extensive
Preprint submitted to Digital Communications and Networks January 22, 2016
simulations using MATLAB, we analyze the effectiveness of these algorithms on network performance for single and multiple gateway scenarios and show that the EZone algorithm tactically controls the topology of the network, thereby maintaining significant service area coverage when compared to the other routing algorithms.
Keywords: Routing, energy efficiency, wireless sensor networks, zones, topology control
1 1. Introduction
2 A wireless sensor network (WSN) is a group of autonomous sensor nodes
3 that are geographically distributed to gather data and monitor events. WSNs
4 are finding increased applicability to the Department of Defense (DoD) in
5 areas specific to surveillance and reconnaissance. A tactical WSN is used in
6 a remote geographic location in order to monitor deployed systems and trig-
7 ger alerts at a command-and-control (C&C) site when certain events occur.
8 Each sensor node in the WSN must have the ability to simultaneously serve
9 as a sensing device and a wireless communication device that can exchange
10 information with nearby nodes [1]. The gateway serves as the destination
11 for a node's packets and is the bridge between the tactical WSN and the
12 backbone infrastructure which includes the C&C site. Because the gateway
13 is a significant component of the WSN architecture, its location must be con-
14 sidered. We focus our attention toward gateway locations on the periphery
15 of the sensor field. For tactical WSNs, we assume that a location on the pe-
16 riphery is more likely to be a safe zone compared to where the sensor nodes
17 are deployed. Our use of safe zone refers to a location where the gateway is
18 outside normal environmental and physical constraints to which sensor nodes
19 may be subjected.
20 In this paper, we investigate two types of tactical WSNs: 1) a single
21 gateway scenario and 2) a multi-gateway scenario. The majority of exist-
22 ing research in WSNs generally includes the perspective of a single gate-
23 way [1, 2, 3, 4, 5]. The few works that study multigateway sensor networks
24 focus on reliable routing, not taking into consideration the energy efficiency
25 requirements of the sensor nodes [6, 7]. Thus, it is important to extend
26 WSN concepts to a multi-gateway framework and identify the resulting per-
27 formance improvements by including an additional gateway.
28 1.1. Deployment Challenges of Tactical WSNs
29 A tactical WSN must operate reliably and increase sensor network cover-
30 age for as long as possible in the absence of human contact. A key challenge
31 in the deployment of tactical WSNs is the limited battery power of each sen-
32 sor node. This has a significant impact on the service life of the network. We
33 define service life of a tactical WSN to be the amount of time that nodes are
34 able to transmit information to the gateway without significant interruption.
35 The service life of the network is contingent upon the network topology. As
36 nodes begin to die out, the remaining live nodes may be disconnected from
37 one another, undermining their ability to communicate with the gateway
se node. For example, if only 20% of the nodes in the network remain alive
39 (i.e., have enough residual energy to use for transmitting, sensing and/or
40 receiving) but they are concentrated within transmission range one another,
41 then communications can still take place for that area. This is the preferred
42 situation. However, with 80% of the nodes dead, the possibility of live nodes
43 residing in areas where they are detached from one another is also possible.
44 While this situation may also occur in a commercially used WSN, the ram-
45 ifications of information not getting to the gateway may not be as severe
46 as in a tactical WSN where important information from the battlespace is
47 being transmitted from the sensor nodes and being used for tactical decision
48 making. Thus, the ability to control the network topology using an effec-
49 tive routing algorithm is essential to ensuring the network remains usable
50 for the most amount of time. In this paper, topology control refers to the
51 ability of the routing algorithm to ensure that nodes with residual energy in
52 one or more areas remain connected to one another and/or the gateway for
53 continued data transfer.
54 1.2. Motivations and Contributions
55 Energy efficient routing is not a new topic in WSN research. Exten-
56 sive studies have been conducted in this area [1, 2, 8, 9, 10, 11, 12, 13, 14].
57 Many of these works offer modifications to already well established WSN
58 routing algorithms. Both [12] and [13] provide algorithms for the modi-
59 fied Low Energy Adaptive Clustering Hierarchy (LEACH) algorithm [2]. In
60 addition, [10], [11], [14] and [15] develop algorithms based on the idea of clus-
61 tering. Clustering is a common hierarchical routing procedure implemented
62 in WSNs. The idea is that energy consumption is reduced by allowing only a
63 select number of nodes, known as cluster heads (CH), to aggregate data from
64 member nodes and transmit to the gateway. A disadvantage with clustering
65 is the CH election process. Depending on the type of procedure used, CHs
66 may be elected such that they reside on the opposite end of the network [16].
67 This situation is common in LEACH where CHs are elected randomly based
68 on a probability model. This means that an elected CH may not be phys-
69 ically close to node members. This then nullifies any energy savings that
70 clustering achieves. There have been various modifications to LEACH in
71 recent years, including improvements to LEACH security [17]. However, the
72 fundamental CH election procedure remains the same, exposing the problem
73 of CH election as mentioned above.
74 There has been some work that has been done on tactical WSNs that serve
75 as a foundation for our work [18, 19]. While [18] and [19] provide architectural
76 constraints for tactical WSN deployment, the routing process and its impact
77 on network topology is not discussed. In [20], the authors develop a cross
78 layer load balancing/routing scheme for tactical WSNs. However, the authors
79 do not provide an in depth analysis on the impact of tactical topology control
80 when using load balancing and routing algorithms.
81 In this paper we show that traditional routing algorithms that are regu-
82 larly used in commercial WSNs have a negative impact on the service life of a
83 tactical WSN because their design is not meant to meet the requirements of
84 tactical WSN applications. More specifically, we extend our work in [20] by
85 showing that established routing algorithms regularly seen in the literature
86 do not effectively control the topology of the network. We aim to provide
87 realistic insights on how an energy efficient routing algorithm can increase
88 service life by tactically controlling the network topology. To the best of
89 our knowledge, this is the first work to provide an extensive analysis of how
90 different routing algorithms impact the operational capability of a tactical
9i WSN.
Our contributions in this paper can be summarized as follows:
• Develop a novel energy efficient zone routing algorithm that tactically controls the network topology. We call this algorithm EZone. We identify performance improvements of EZone and compare it to the following established routing techniques: 1) Direct Routing, 2) Minimum Transmission Energy (MTE), 3) Low Energy Adaptive Clustering Hierarchy (LEACH), and 4) Zone routing. We also identify performance improvements of adding an additional gateway to these algorithms.
• As sensor-node battery levels are depleted and nodes subsequently die out, we show how EZone affects the topology of live nodes and dead nodes in the sensor field and how this affects the continuous service cov-
erage throughout the sensor field. We compare EZone service life with
104 the network service life obtained using each of the four routing algo-
105 rithms mentioned above and show EZone's ability to tactically provide
106 continuous service.
107 The remainder of this paper is organized as follows. In Section 2 we
108 discuss the models implemented at each layer of the tactical WSN protocol
109 stack. In Section 3 we discuss the EZone algorithm and its implementation.
110 We also describe the four traditional routing algorithms that EZone is com-
111 pared to. We provide our simulations and analysis of the results in Section 4.
112 We conclude the paper in Section 5.
113 2. Tactical WSN Protocol Stack Implementation
114 We implemented the following models into each layer of the protocol
115 stack.
116 Physical layer: All nodes in our simulations begin with a starting energy
117 level of 0.5 Joules (J). This is a value commonly used in the literature because
118 it provides small enough energy to quickly see the effects of the varying
119 algorithms involved yet it provides enough energy to demonstrate node life
120 longevity by making algorithmic improvements.
121 The physical model relates the amount of energy a sensor node consumes
122 during transmit and receive operations. Several power energy consumption
123 models exist in the literature [21, 22, 23, 24]. Each model presents a differ-
124 ent way of calculating total energy consumption for different sensor nodes.
125 We chose to utilize a first order power amplifier and sensor model for sim-
126 plicity and because it is more prevalently used in the literature [2, 25, 26].
127 This model assigns an energy cost-per-bit to collect, transmit and receive
128 information. It considers direct path and multi-path wireless signal propa-
129 gation theory to identify the amount of information required to transmit one
130 bit of information over a certain distance between nodes while guaranteeing
131 adequate signal-to-noise (SNR) ratio at the receiving node. We utilize the
132 first order radio energy model to relate the energy expended to send and
133 receive an L-bit message over a distance d when considering direct path and
134 multi-path propagation [1, 2, 16, 27].
135 The energy expended in the transmit electronics for free space (direct
136 path) propagation, ETx-fs, is described by
ErX-fs(L d) = ETx-elec(L) +ETx-amp(Lj d) = EelecL + EfsLd? (1)
137 and for multipath propagation by
ETx-mp(Lj d) =ETx-elec(L) + ETx-amp(Lj d) = EelecL + empLd4 (2)
138 where Edec corresponds to the energy per bit required in transmit and receive
139 electronics to process the information, ETx-amp is electrical energy required to
140 transmit an L-bit message over a distance d, and £fs and smp are constants
141 corresponding to the energy per bit required in the transmit amplifier to
142 transmit an L-bit message with adequate SNR over a distance d2 and d4 for
143 free space and multi-path propagation modes, respectively.
144 The energy expended to receive the L-bit message in the receive electron-
145 ics is described by
ERx (L) = EelecL (3)
146 The corresponding values from Eqs. 1-3 for the amplifiers and electronics
147 used in our subsequent simulations are described in Table 1.
Table 1: Radio Energy Dissipation Parameters [4]
Constant Value
Transmit and Receive Electronics, Eelec 50 nJ/bit
Transmit Amplifier, free space propagation, Sfs 10 pJ/bit/m2 Transmit Amplifier, multi-path propagation, emp 0.0013 pJ/bit/m4
148 MAC layer: We simulate the MAC layer simply through the performance
149 of transmission rounds. Each simulation begins at round one and ends when
150 the last node dies. During each round, each node in the WSN sends an L
151 bit packet to the gateway. We implement a Time-Division Multiple Access
152 (TDMA) scheme that assigns each node in the WSN a timeslot during each
153 round. The node transmits information to the gateway during the timeslot.
154 With the clustering and zoning algorithms, we assume the MAC process is
155 similar to that described for LEACH, in which CHs are assigned a TDMA
156 timeslot for transmission to the gateway and CHs are assigned code-division
157 multiple access (CDMA) schemes for intra-cluster communications to prevent
158 interference with other clusters/zones.
159 Network layer: There are a variety of routing algorithms applied to the
160 network layer in the literature, some of which are described in [3, 5, 4, 28,
161 29]. We implement several traditional and established routing algorithms
162 observed in the literature. We also develop and implement our own energy
163 efficient routing algorithm (EZone). The routing algorithms we implement
164 are direct, MTE, LEACH, zone, and EZone. We will discuss these algorithms
165 in further detail in Section 3.
166 Transport layer: Our transport layer implements User Datagram Protocol
167 (UDP).
168 Application layer: Our application layer implements two strategies: 1)
169 use of a traffic generator, and 2) use of a data aggregation technique. The
170 traffic generator of each node generates a 2000 bit data message during each
171 round for transmission to the gateway. Data aggregation is used only for the
172 clustering and zone routing algorithms and the CH is the only node that can
173 perform data aggregation. The CH receives all the messages from nodes in
174 the cluster. It then includes its own message, compresses all the messages
175 into one 2000-bit message, and transmits the compressed message to the
176 gateway at the end of each round.
177 Data aggregation requires energy to perform the signal compression, which
178 must be accounted for. We adopt a similar technique used in the literature,
179 which applies an energy cost to the data aggregator for the task of aggregat-
180 ing all the data during a round. A data aggregation constant, EDA (event
181 oriented data aggregation) is used to account for the energy to compress
182 messages into one final L=2000 bit message. The data aggregation constant
183 used in our scenarios is consistent with the literature (EDA = 5 nJ/bit) and
184 results in an aggregation cost of EDA x L [1, 2, 27, 30, 31].
185 3. Routing Algorithms for Tactical WSNs: Traditional vs EZone
186 In this section, we describe the five routing algorithms that were simu-
187 lated: Direct, MTE, LEACH, Zone, and EZone.
188 3.1. Traditional Routing Algorithms: Direct, MTE, LEACH and Zone
189 Direct transmission to the gateway involves each node sending a packet
190 to the gateway directly without using any other nodes along the way. During
191 each round, the Euclidean distance is calculated between the node and the
192 gateway. The distance along with the transmit amplifier parameters given
193 in Table 1 is used to determine the propagation mechanism [16]. The node's
194 energy is decremented in proportion to the required energy for packet trans-
195 mission to the gateway.
196 In MTE routing we minimize the propagation distance to the gateway in
197 order to produce a route that minimizes the overall sensor energy depletion
198 rate. We utilize propagation distance as our link cost parameter to input
199 into the MTE algorithm. We use Dijkstra's algorithm to generate our MTE
200 routes. In MTE routing, the node closest to the gateway is always chosen to
201 be included in the route. This node is known as the hot node. Since the hot
202 node is the relay point between the gateway and all traffic from other nodes,
203 it is overwhelmed with traffic during each round and dies quickly. Another
204 hot node is then immediately chosen. This hot node concept in MTE routing
205 causes nodes that are closest to the gateway to die out first.
206 The LEACH algorithm is a well-known clustering algorithm developed
207 specifically for WSNs. LEACH routing elects one or more CHs and nodes
208 associate with the nearest CH. The role of CH is rotated among the nodes
209 in the following way: Each node picks a random number between zero and
210 one. Each node also computes a threshold number (Tn), which is a number
211 between zero and one and is proportional to the current round. The proba-
212 bility for any node to serve as a CH is denoted as p. If a node has been a CH
213 in the last ^ rounds, it is excluded from being a CH during the round. Oth-
214 erwise, if the temporary random number is less than Tn, the node is elected
215 as a CH during the round. The desired probability for a node to be chosen
216 as a CH is an input to the algorithm and must be specified. The original
217 authors of LEACH performed analysis to determine the optimum value for
218 p to be 0.05 [1]. Each node transmits its data message to its CH. Each CH
219 collects all the messages of its nodes and retransmits them collectively to the
220 gateway. This process repeats during subsequent rounds until all nodes have
221 died.
222 Zone clustering appears less frequently in the literature as compared to
223 LEACH. However, for a tactical network, it may be a preferred routing al-
224 gorithm because the user can specify how zones are characterized for the
225 network. The general methods used for the zone routing algorithm are based
226 on techniques described in [8]. In [8], the authors utilize a sensor field com-
227 prised of homogenous zones. Partitioning the network in zones essentially
228 creates several smaller WSNs that all utilize the same gateway. A sensor in
229 each zone has a probability p of becoming a CH during each round. The
230 probability p is determined to be relative to the number of nodes in the zone:
231 p = --r-f 1 i—---. The zone clustering algorithm divides the
^ (number of nodes in zone) ° °
232 sensor field into z equal zones. Equal zones span along the Cartesian x-axis
233 to create z vertical rectangular zones. We use five zones in our simulations.
234 Five zones were chosen to provide a comparison with the LEACH algorithm.
235 Recall that in the LEACH algorithm, the probability of any node being cho-
236 sen a CH is p=0.05. Thus, in a 100 node network, we would have five CHs.
237 To ensure there are five CHs for our zone clustering algorithm, we must have
238 five zones and each zone is only allowed to have one CH.
239 During each round, the set of live nodes for each zone is identified, and
240 the CH is chosen based on a random assignment from this set. Each node
241 in the zone then transmits its L-bit packet to the zone's CH and its energy
242 is decremented according to our radio energy model. The CH for the zone
243 then aggregates all the messages from the nodes in the zone and transmits
244 the aggregated message to the gateway.
245 For all four routing algorithms discussed in this section, the multigateway
246 scenarios operate the same way as described, except the gateway that is
247 closest to each node in terms of Euclidean distance is chosen to receive data.
248 3.2. EZone: Zone Clustering with Energy Efficient Cluster Head Selection
249 The zone clustering case described in Section 3.1 chooses the CH for each
250 zone randomly. A clustering algorithm that partitions nodes into specific
251 zones is an energy saving technique when compared to the LEACH algorithm
252 because there is a lower maximum distance that any node must transmit to
253 reach its CH. Zone routing guarantees a nearby CH in the zone as compared
254 to that of LEACH. In LEACH the nearest CH may be on the other side of
255 the network since the criteria for a node to be elected as a CH may have only
256 been met randomly on the other side of the field [16].
257 There are significant differences in energy distribution of the nodes in the
258 network. The differences in energy levels across the WSN cause some nodes
259 to die out earlier and some nodes to die out later. Therefore, in the EZone
260 algorithm, we modify the CH election criteria in the following way: in any
261 given round, if the highest energy node is chosen to be the CH, individual
262 node energy depletion rates are decreased allowing battery levels in any zone
263 to deplete at a uniform rate.
264 To accomplish this strategy, the zone routing algorithm is revised. Instead
265 of randomly choosing the CH from the live nodes in the zone, we choose the
266 CH that has the maximum energy level in the zone. Based on this election
267 criteria, nodes that are in a more preferred location (a location that decreases
268 energy depletion rate such as locations closer to the gateway) are chosen to
269 be the CH for the zone more than those in a less preferred location (a location
270 farther away from the gateway).
271 The EZone algorithm is executed in three phases: 1) network setup; 2) CH
272 election for each zone; and 3) packet transmission from CH to gateway. The
273 network setup phase creates the WSN and partitions the network into the
274 required number of zones. The number of zones is based on user requirements
275 and application scenario. We use 5 zones to facilitate comparison with the
276 zone routing algorithm using random CH election given in [8]. Partitioning
277 the network into zones effectively creates several smaller WSNs that all utilize
278 the same gateway. The zone assigned to any node is based on the node's x-
279 coordinate in the network field. Once all nodes are assigned to a zone, we
280 begin the simulation at round one. In each round, the set of live nodes for
281 each zone is identified and the CH is elected based on highest node energy.
282 Electing the highest energy node to be the CH during each round in each
283 zone requires additional processing by the gateway to perform CH election.
284 In order for the gateway to make an effective CH choice for each zone, it
285 must be aware of all the alive nodes in each zone and the residual energy
286 (remaining energy) of each alive node. Each node in the zone maintains a
287 power meter that is used to maintain node residual power. Each alive node in
288 each zone decrements its power meter each time it transmits a packet to the
289 clusterhead of that zone. The decrement is based on the radio energy model
290 given in Eqs. 1 and 2. All alive nodes communicate with the gateway during
291 the start of each round. During round 1 only, the gateway chooses the CH
292 randomly, similar to [8]. The reason the CH is chosen randomly for round 1 is
293 because it is assumed that at the start of the algorithm all nodes have equal
294 energy and thus any node can be the CH. The CH of round 1 transmits
295 an aggregated packet to the gateway. The aggregated packet includes the
296 residual node energies for each alive node in the zone (including the current
297 CH) in the packet header. The gateway uses the residual power values to
298 choose the CH for each zone for the next simulation round. The node with
299 the highest residual energy is chosen by the gateway to be the CH for the
300 subsequent round. The CH choice is then broadcast back to each zone. Once
301 a CH is elected by the gateway, the CH maintains its power meter by 1)
302 decrementing the energy required to aggregate and send the packet for that
303 round to the gateway based on the radio energy model and 2) decrementing
304 the energy cost for the CH to receive packets from nodes in its zone. The
305 decrement is calculated based on Eq. 3.
306 This CH selection based on highest node residual energy is a minor ad-
307 justment from the traditional zone routing algorithm but it has a significant
308 effect on the service life and network topology of a tactical WSN, as will be
309 shown in Section 4.
310 4. Simulations and Result Analysis
311 In our simulations, sensors and gateways are all placed on a Cartesian grid
312 with axes x and y. Our simulations and analysis involve a grid of 100 sensors
313 such that each sensor's x and y coordinate is modeled as a uniformly dis-
314 tributed random variable between 0 and 50 meters (m). The single gateway
315 scenario employs the gateway at (x,y) = (25m, -100m). In the multigateway
316 scenario an additional gateway is placed at the position (25m, 150m). In
317 the subsequent figures, gateways are displayed as solid green nodes and live
318 nodes are represented by a blue outline circle. We show the perimeter and
319 zone fields as solid red lines. All nodes have a starting energy of 0.5 J, ex-
320 cept the gateway(s) which is assumed to have unlimited energy (they are not
321 energy constrained). The traffic routed in the network is generated using a
322 constant bit rate (CBR). The size of each message (L) is 2000 bits.
323 All our simulations assume that each node is within wireless transmission
324 range of the gateway which also means that each node is within communi-
325 cation range of any other node in the WSN. We make this assumption in
326 order to simplify the simulation scenario. This simplification makes it easier
327 to analyze node-die out statistics and network topology characteristics for
328 each routing algorithm executed. In our future work, we will loosen this
329 assumption. All our simulations were executed in MATLAB.
330 4-1- Topology Control Analysis of Routing Algorithms: Traditional vs EZone
331 In every round we generate several plots to characterize energy consump-
332 tion and the distribution of live and dead nodes in the network. We produce
333 three plots during each round. The first plot is a bar plot that provides the
334 energy of each node from 1 to 100 where node 1 is the closest node to x=0
335 (the y-axis) and node 100 is on the other side of the sensor field closest to the
336 line x=50 m. The second plot is a three-dimensional energy stem plot where
337 each stem is located in the position of the node in the field, and the height
338 of the stem represents the amount of residual battery energy available. The
339 energy stem plot is green, and the elevation (energy level) decreases during
340 each round, corresponding to energy consumption. When the stem reaches
341 zero energy (the floor), the green bubble changes to red to indicate the node
342 has died. The final plot is an overview of the sensor field topology including
343 the gateway during a particular round. We refer to the first plot as the en-
344 ergy bar plot, the second plot as the energy stem plot and the third plot as
345 the node distribution plot. The node distribution plot shows live nodes as a
346 circle with a blue outline and dead nodes as solid red bubbles. The node dis-
347 tribution plot also contains the round from which all three plots are drawn.
348 The energy bar and stem plots are stacked on top of each other on the left
349 hand side of the figures, and the node distribution plot is on the right side of
350 the figures. For each simulation, this is plotted four times corresponding to
351 the round the first node dies and the round that 10%, 50%, and 80% of nodes
352 have died. To constrain the length of this paper, we provide the plots for
353 80% of nodes dead to illustrate the operational mechanism of the direct and
354 MTE routing algorithms. We provide the plots for 10% and 50% of nodes
355 dead to illustrate LEACH and generic zone routing. We provide plots for
356 1 0%, 50%, and 80% of nodes dead for the EZone routing algorithm. Due to
357 space, these plots are graphically shown for the single gateway scenario only
358 (in Section 4.2 we provide further discussion on the multigateway scenarios
359 by providing ensemble graphs that show the performance of each routing
360 algorithm in multigateway scenarios). The results depicted in Figs. 1-9 are
361 simulations of one specific topological configuration.
362 Direct transmission: The plot for 80% of nodes dead is shown in Fig. 1.
363 The energy stem plot demonstrates that nodes closest to the gateway remain
364 in service longer than nodes farther from the gateway because our physical
365 layer depletes energy proportional to distance.
see MTE with Dijkstra routing: The plot for 80% of nodes dead is shown
367 in Fig. 2. The energy stem plot and node distribution plots demonstrate that
368 nodes closest to the gateway die out first and then fan out as subsequent live
369 nodes closest to the gateway become the hot nodes. As expected, this quickly
370 eliminates service coverage in those areas. Nodes that are farthest away from
371 the gateway are not used by their peers as frequently for routing, thus their
372 energy is preserved. This creates a large energy variance and the quickest
Figure 1: Illustration of a single gateway tactical WSN for direct routing. The network topology when 80% of the nodes are dead is on the right of each subplot, the energy stem plot is shown in the lower left and the energy bar plot is on the upper left of the subplots
OiS S 03 ? 035 c
q] 0? 0.15 01
20 № BO 80 100
Node Number
OO t> o
•" * V' * "
* ■ m *
> . ^ / - --
Round 354
Figure 2: MTE routing in a single gateway tactical WSN illustrating the network topology when 80% of the nodes are dead
373 die out for all results collected in this paper (variance results produced by
374 each routing algorithm are further examined in the next section).
375 LEACH: The 10% and 50% of nodes dead are plotted in Figs. 3 and 4,
376 respectively. The CHs are indicated by a blue asterisk that fills the nodes.
377 Our display of CHs involves one caveat for LEACH and zone routing algo-
378 rithms. In some cases, the CH asterisk indicator is plotted over with a solid
379 red circle because its energy was fully depleted in its last round as the CH.
380 Our plots are drawn at the end of each round; thus, if a node is dead and
381 it was the CH during the round, it is depicted as a dead node. The en-
382 ergy stem plot and node distribution plots in both figures demonstrate that
383 nodes die out starting in the middle of the network and progress out. From
384 this outward progression, nodes toward the top of the network die out more
385 quickly than nodes at the bottom of the sensor field because nodes at the top
386 use more energy to transmit a cluster's payload to the gateway during the
387 random times they are selected as the CH. Nodes at the center of the field
388 start to die out first as a result of LEACH's mechanism for determining CHs
389 and cluster assignments at each round. Thus, this shows that LEACH inef-
390 ficiently partitions the sensor field with CHs, without regard to any spatial
391 arrangement.
392 Zone Routing: The plots for 10% and 50% of nodes dead are shown in
393 Figs. 5 and 6, respectively. The energy stem plot and node distribution plots
394 demonstrate much more uniform energy depletion as compared to all the
Figure 3: LEACH routing in a single gateway tactical WSN illustrating the network topology when 10% of the nodes are dead
05 0.45
036 5 0.3
ill IL. IlLl.U
Node Number
JO ^ _-^i-»*" JO
Hound 1990
ZO 30 40
x-Grid-Axis (it \
Figure 4: LEACH routing in a single gateway tactical WSN illustrating the network topology when 50% of the nodes are dead
395 other algorithms tested thus far. Since any node can be randomly selected
396 to be a CH more than any other node in the network, this creates a random
397 mode for nodes to die out. Zones in the node distribution plots die out
398 consistently with no one zone dying out earlier than another zone. We found
399 that when the zone algorithm is run, the first node dies at round 1649,
400 providing the longest service life of 100% of nodes alive of all algorithms
401 tested thus far. This will be further discussed in the next section.
Round 1821
x-Grid-Axis (m)
Figure 5: Zone routing in a single gateway tactical WSN illustrating the network topology when 10% of the nodes are dead
EZone: We provide plots for the EZone routing algorithm when 10%,
403 50% and 80% of the nodes die. These plots are shown in Figs. 7-9. The
404 energy stem plot and node distribution plots demonstrate that zones die out
405 from the outer zones in the sensor network, progressing toward the center.
406 By restricting CH election criteria to choose the highest energy node in a
3 03 i? 0.21 £ 01 015
JiiijI.Lj il.niilil ..lli
Ned« Number
1 0 a —1—*j ■-
» ■ a ■ • 0 ' ° ,» 0
° t* h ö • o 0 < a o m
■ s C o ra 0
Round 2022
Figure 6: Zone routing in a single gateway tactical WSN illustrating the network topology when 50% of the nodes are dead
407 zone, the energy level of all nodes in a common zone are uniformly preserved
408 throughout the simulation and thus nodes die out evenly.
409 Notice in Fig. 9, the middle zone is 100% still in service even when all
410 nodes in the other four zones are dead. This is a valuable consequence of the
411 EZone algorithm. Even though 80% of the nodes are dead, the fact that one
412 zone is still fully operational is significant for tactical operations; data sensed
413 and transmitted within that middle zone will be sent to the gateway. Nodes
414 within that middle zone have neighboring nodes that are still alive through
415 which to transmit data. This ensures that the network remains valuable from
416 a tactical perspective.
Figure 7: EZone routing in a single gateway tactical WSN illustrating the network topology when 10% of the nodes are dead
x-Grid-Axis (m)
Figure 8: EZone routing in a single gateway tactical WSN illustrating the network topology when 50% of the nodes are dead
417 4-2. Comparisons of the Algorithms Based on Energy Consumption
418 We plotted the total tactical WSN system energy level during each trans-
419 mission round (Fig. 10), the energy variance that resulted from the distribu-
420 tion of individual node battery levels (Fig. 11) and the number of live nodes
421 during each round (Fig. 12). We visually observed how nodes geographically
422 die out throughout the simulation. The results shown in Figs. 10-12 were
423 obtained when 5000 different network topologies were simulated. Each rout-
424 ing algorithm was executed on each of the 5000 topologies and the average
425 results are depicted in the figures. In each legend of Figs. 10-12, an S after
426 an algorithm name refers to the single gateway scenario, and M refers to the
x-Grid-Axis (m)
Figure 9: EZone routing in a single gateway tactical WSN illustrating the network topology when 80% of the nodes are dead
427 multigateway scenario. The variance is given by the following equation.
428 where E is the random variable for energy, n represents the number of live
429 nodes in the round, ei is the energy of the ith live node in the round and ^
430 is the mean energy for the round.
431 The clustering algorithms dramatically outperformed the MTE and di-
432 rect routing algorithms. This is a result of rotating and distributing the high
433 energy role of nodes that perform long-range transmission and data aggrega-
434 tion. The single and multigateway clustering algorithms generally displayed
0 500 'JO.i 1W0 2 BOO 2500
Figure 10: Total tactical WSN system energy versus transmission round for direct, MTE, LEACH, zone and EZone routing algorithms in single gateway and multigateway scenarios
Figure 11: Energy variance versus transmission round for direct, MTE, LEACH, zone and EZone routing algorithms in single gateway and multigateway scenarios
Figure 12: Total number of live nodes versus transmission round for direct, MTE, LEACH, zone and EZone routing algorithms in single gateway and multigateway scenarios
435 similar energy depletion rates that are illustrated in the linear regions of
436 Fig. 10. The clustering algorithms decrease the energy variance of the tacti-
437 cal WSN, and our energy efficient zone routing algorithm, EZone, provided an
438 indistinguishable flat variance plot compared to other algorithms, as shown
439 in Fig. 11. In Fig. 12 it can be seen that EZone increased the time when all
440 nodes are alive, with the single gateway EZone simulation outperforming all
441 other algorithms.
442 Our EZone algorithm outperformed all other algorithms from a topology
443 perspective during node die out as well. While other algorithms created
444 a pattern for die out, our energy efficient algorithm kept all nodes in one
445 area/zone alive for the longest time possible. Node die out of other routing
446 algorithms occurred in an unfavorable fashion. For example, in the direct
447 case, live nodes farther from the gateway died first since their energy is
448 depleted proportional to their distance from the gateway. As a result, areas
449 farthest from the gateway lost service first, while areas closest to the gateway
450 remained in service longest. In MTE routing the nodes closest to the gateway
451 died first. The LEACH algorithm inefficiently creates clusters that cause the
452 network to die out starting in the center of the sensor field and progressing
453 radially outward. As a result of this die out mechanism, we lose coverage in
454 the middle of the sensor field first. These die out mechanisms warrant the
455 choice of EZone for a tactical WSN since it preserves 100% network coverage
456 the longest time for specific zones.
457 The addition of another gateway was most significant in the direct and
458 MTE algorithms as the energy variance is lowered by approximately 50 per-
459 cent. This can be seen in Fig. 11. The energy variance of the zone routing
460 algorithms were both lower than LEACH, with the single gateway scenarios
461 performing better than LEACH in a multigateway configuration. In Fig. 12
462 we can see that the EZone-S and EZone-M plots keep 100% of the nodes alive
463 until approximately rounds 2100 and 2200, respectively, and then there is a
464 significant drop in the number of nodes alive. In comparison, the LEACH-S
465 and LEACH-M plots show that 100% of the nodes are alive until approxi-
466 mately round 1800 but there is a more gradual drop off in the number of
467 nodes alive. This supports our assertion that EZone offers the most time
468 with all nodes alive whereas LEACH offers the most time with at least one
469 node alive.
470 This is an important distinction, particularly when dealing with tactical
471 operations. While having at least one node alive may be useful in commercial
472 WSNs, it is not practical for WSNs deployed for tactical operations. With
473 one node alive, there is very little that can be done, unless the node is very
474 close to the gateway. Even if more than one node is alive, the possibility of
475 having nodes alive in different areas of the network that are not connected
476 is also not helpful in tactical operations. Thus, EZone's ability to keep all
477 nodes alive for a longer period of time than LEACH combined with its ability
478 to keep all nodes within one zone alive (see Fig. 9) makes it more applicable
479 for use in tactical networks.
480 5. Conclusion
481 Tactical WSNs are used by the military to obtain information about
482 ground situational awareness. This information facilitates tactical decision
483 making. In order to increase service life in various areas of a tactical WSN,
484 we need to control the network topology such that nodes with residual en-
485 ergy are used and maintained for continued communication. We develop an
486 energy efficient zone routing algorithm, called EZone, to tactically control
487 the network topology by providing 100% service life of all nodes in specific
488 zones/areas of the network for a longer period of time when compared to
489 other energy aware routing schemes, in particular LEACH. Our EZone algo-
490 rithm offers the best opportunity to extend tactical WSN service life while
491 maintaining tactical control of the network in both single and multigateway
492 configurations. It produced the least variance in energy distribution at any
493 round and smartly balanced cluster and node traffic balancing to decrease
494 energy consumption.
495 References
496 [1] W.R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, "Energy-
497 efficient communication protocol for wireless microsensor networks," in
498 Proc. of Hawaii International Conference on System Sciences (HICSS),
499 2000, pp. 1-10.
500 [2] W.R Heinzelman, A. Chandrakasan, and H. Balakrishnan, "An appli-
501 cation specific protocol architecture for wireless microsensor networks,"
502 IEEE Transactions on Wireless Communications, vol. 1, no. 4, pp. 660503 670, October 2002.
504 [3] J. Hao and H. Mouftah, "Routing protocols for duty cycled wireless
505 sensor networks: A survey," IEEE Communications Magazine, vol. 50,
506 no. 12, pp. 116-123, December 2012.
507 [4] C. Cirstea, "Energy efficient routing protocols for wireless sensor net-
508 works: A survey," in Proc. of IEEE International Symposium for Design
509 and Technology in Electronic Packaging (SIITME), 2011, pp. 277-282.
510 [5] D. Goyal and M.R. Tripathy, "Routing protocols in wireless sensor
511 networks: A survey," in Proc. of IEEE Advanced Computing and Com-
512 munication Technologies (ACCT), 2012, pp. 474-480.
513 [6] P. Thulasiraman, S. Ramasubramaniam, and M. Krunz, "Disjoint mul-
514 tipath routing in dual homing networks using colored trees," in Proc. of
515 IEEE Global Communications Conference (Globecom), 2006, pp. 1-5.
516 [7] P. Thulasiraman, S. Ramasubramaniam, and M. Krunz, "Disjoint mul-
517 tipath routing to two distinct drains in a multi-drain sensor network," in
518 Proc. of IEEE International Conference on Computer Communications
519 (INFOCOM), 2007, pp. 643-651.
520 [8] S.K. Singh, M.P. Singh, and D.K. Singh, "Energy efficient homogenous
521 clustering algorithm for wireles sensor networks," International Journal
522 of Wireless Mobile Networks, vol. 2, no. 3, pp. 49-61, 2010.
523 [9] F. Bouabdallah, N. Bouabdallah, and R. Boutaba, "Load-balanced rout-
524 ing scheme for energy-efficient wireless sensor networks," in Proc. of
525 IEEE Global Communications Conference (Globecom), 2008, pp. 1-6.
526 [10] A.S.K. Mammu, A. Sharma, U. Hernandez-Jayo, and N. Sainz, "A
527 novel cluster-based energy efficient routing in wireless sensor networks,"
528 in Proc. of IEEE Advanced Information Networking and Applications
529 (AINA), 2013, pp. 41-47.
530 [11] R. Ibrahim, H. Quang-Duang, and T. Le-Ngoc, "An energy-efficient and
531 load-balancing cluster-based routing algorithm for csma-based wireless
532 sensor networks," in Proc. of IEEE Vehicular Technology Conference
533 (VTC), 2013, pp. 1-5.
534 [12] M. Rangchi and H. Bakhshi, "A new energy efficient routing algorithm
535 based on load balancing for wireless sensor networks," in Proc. of Inter-
536 national Symposium of Telecommunications (IST), 2014, pp. 1201-1205.
537 [13] D. Kumar, "Performance analysis of energy efficient clustering proto-
538 cols for maximising lifetime of wireless sensor networks," IET Wireless
539 Sensor Networks, vol. 4, no. 1, pp. 9-16, 2013.
540 [14] S. Soro and W. Heinzelmann, "Cluster head election techniques for
541 coverage preservation in wireless sensor networks," Ad Hoc Networks
542 (Elsevier), vol. 7, no. 5, pp. 955-972, 2009.
543 [15] J.-Y. Lee, K. Jung, H. Jung, and D. Lee, "Improving the energy effi-
544 ciency of a cluster head election for wireless sensor networks," Interna-
545 tional Journal of Distributed Sensor Networks, vol. 2014, pp. 1-6, 2014.
546 [16] K.A. White, "Tactical network load balancing in multigateway wireless
547 sensor networks," M.S. thesis, Naval Postgraduate School, Monterey,
548 CA, USA, December 2013.
549 [17] M. Alshowkan, K. Elleithy, and H. Alhassan, "Ls-leach: A new secure
550 and energy efficient routing protocol for wireless sensor networks," in
551 Proc. of IEEE/ACM 17th International Symposium on Distributed Sim-
552 ulation and Real Time Applications (DS-RT), 2013, pp. 215-220.
553 [18] H. Song, S. Lee, S. Lee, and H. Lee, "6lowpan-based tactical wireless
554 sensor network architecture for remote large-scale random deployment
555 scenarios," in Proc. of IEEE Military Communications Conference (Mil-
556 Com), 2009, pp. 1-7.
557 [19] S. Lee, S. Lee, H. Song, and H. Lee, "Wireless sensor network design
558 for tactical military applications: remote large-scale environments," in
559 Proc. of IEEE Military Communications Conference (MilCom), 2009,
560 pp. 911-917.
561 [20] K.A. White and P. Thulasiraman, "Energy efficient cross layer load bal-
562 ancing in tactical multigateway wireless sensor networks," in Proc. of
563 IEEE International Inter-Disciplinary Conference on Cognitive Method
564 in Situation Awareness and Decision Support (CogSIMA), 2015, pp.
565 193-199.
566 [21] E. Shih, N. Ickes, R. Min, A. Sinha, A. Wang, and A. Chandrakasan,
567 "Phyiscal layer driven protocol and algorithm design for energy efficient
568 wireless sensor networks," in Proc. of ACM MobiCom, 2001, pp. 272569 287.
570 [22] J. Hill, R. Szewczyk, A. Woo, S. Hollar, D. Culler, and K. Pister, "Sys-
571 tem architecture directions for networked sensors," ACM SIGPLAN
572 Notices, vol. 35, no. 11, pp. 93-104, 2000.
573 [23] J. Polastre, J. Hill, and D. Culler, "Versatile low power media access for
574 wireless sensor networks," in Proc. of ACM SenSys, 2004, pp. 95-107.
575 [24] V. Shnayder, M. Hempstead, B. Chen, G. Allen, and M. Welsh, "Simu-
576 lating the power consumption of largescale sensor network applications,"
577 in Proc. of ACM SenSys, 2004, pp. 188-200.
578 [25] I. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, "Wireless
579 sensor networks: A survey," Computer Networks, vol. 38, no. 4, pp.
580 393-422, March 2002.
581 [26] "Tinyos," http://www.tinyos.net.
582 [27] G. Smaragdakis, I. Matta, and A. Bestavros, "Sep: A stable election
583 protocol for clustered heterogeneous wireless sensor networks," Tech.
584 Rep., Boston University, 2004.
585 [28] A.A. Abbasi and M. Younis, "A survey on clustering algorithms for
586 wireless sensor networks," Computer Communications, vol. 30, no. 14587 1 5, pp. 2826-2841, 2007.
588 [29] K. Akkaya and M. Younis, "A survey on routing protocols for wireless
589 sensor networks," Ad Hoc Networks, vol. 3, no. 3, pp. 325-349, 2005.
590 [30] C. Intanagonwiwat, R. Govindan, and D. Estrin, "Directed diffusion:
591 a scalable and robust communication paradigm for sensor networks,"
592 in Proc. of ACM International Conference on Mobile Computing and
593 Networking (MobiCom), 2000, pp. 56-67.
594 [31] F. Ye, H. Luo, J. Cheng, S. Lu, and L. Zhang, "A two-tier data dis-
595 semination model for large-scale wireless sensor networks," in Proc. of
596 ACM International Conference on Mobile Computing and Networking
597 (MobiCom), 2012, pp. 148-159.
