sensors Article EDOVE: Energy an Depth Variance-Base Opportunistic Voi Avoiance Scheme for Unerwater Acoustic Sensor Networks Safar Hussain Bouk 1, *, Sye Hassan Ahme 2, Kyung-Joon Park 1 an Yongsoon Eun 1 1 Department of Information an Communication Engineering, DGIST, Daegu 42988, Korea; kjp@gist.ac.kr (K.-J.P.); yeun@gist.ac.kr (Y.E.) 2 School of Computer Science an Engineering, Kyungpook National University, Daegu 41566, Korea; s.h.ahme@ieee.org * Corresponence: bouk@gist.ac.kr; Tel.: +82-10-3097-0786 Receive: 17 August 2017; Accepte: 22 September 2017 ; Publishe: 26 September 2017 Abstract: Unerwater Acoustic Sensor Network (UASN) comes with intrinsic constraints because it is eploye in the aquatic environment an uses the acoustic signals to communicate. The examples of those constraints are long propagation elay, very limite banwith, high energy cost for transmission, very high signal attenuation, costly eployment an battery replacement, an so forth. Therefore, the routing schemes for UASN must take into account those characteristics to achieve energy fairness, avoi energy holes, an improve the network lifetime. The epth base forwaring schemes in literature use noe s epth information to forwar ata towars the sink. They minimize the ata packet uplication by employing the holing time strategy. However, to avoi voi holes in the network, they use two hop noe proximity information. In this paper, we propose the Energy an Depth variance-base Opportunistic Voi avoiance (EDOVE) scheme to gain energy balancing an voi avoiance in the network. EDOVE consiers not only the epth parameter, but also the normalize resiual energy of the one-hop noes an the normalize epth variance of the secon hop neighbors. Hence, it avois the voi regions as well as balances the network energy an increases the network lifetime. The simulation results show that the EDOVE gains more than 15% packet elivery ratio, propagates 50% less copies of ata packet, consumes less energy, an has more lifetime than the state of the art forwaring schemes. Keywors: voi avoiance; epth variance; normalize resiual energy; energy balancing 1. Introuction Unerwater acoustic sensor networks (UASNs) have gaine a lot of attention from the research an inustrial community ue to their vast application areas. The applications of USANs range from marine ecosystem monitoring, seismic activity, offshore exploration monitoring an management, water pollution, to coastal area surveillance. All of those applications require continuous sensing of the aquatic parameters an relaying these sense parameters towars the onshore station. Sensing an especially communication in the aquatic environment is a quite challenging compare to the terrestrial sensing networks an coul not be realize without UASNs. Main elements of the UASNs are the sensing noes an the sink noes. The sensing noes or simply calle sensors have the sensing an communication capabilities, they operate on the battery power, an usually, they rest at the bottom of the sea an/or anchore to the seabe. Normally, the anchore noes in UASNs are the noes that know their location information [1]. However, here the concept of anchore noe is the noe that is attache with the cor, which is connecte to the anchor. In aition to the sensing an communication, these sensor noes can store small size Sensors 2017, 17, 2212; oi:10.3390/s17102212 www.mpi.com/journal/sensors
Sensors 2017, 17, 2212 2 of 25 contents ue to the limite storage capacity of the noe. Sensor noes in the aquatic environment can communicate using the acoustic [2,3], electromagnetic/raio [4,5], or optical communication moes [6,7]. Sensor noes commonly use the acoustic communication moe. However, the acoustic communication moe inherits some peculiar characteristics. Acoustic communication has a very long communication range, a very long propagation elay, high absorption an path loss, high energy requirement for the transmission an reception of the acoustic signal, limite banwith ue to high error rate, an so forth [8,9]. On the other han, the sink noes or simply Sinks float over the sea surface. Sinks use multiple interfaces, one interface to communicate with the unerwater eploye noes (e.g., acoustic communication moe) an the other interface to connect with the onshore monitoring an control center (e.g., raio communication moe). The sensor noes forwar or route the sense ata towars the Sink which further relays this ata to the onshore monitoring stations. The generic UASN scenario is shown in Figure 1, where multiple sinks are floating over the sea surface. The unerwater acoustic sensors are eploye uner water with the acoustic transmission range of t r. In that scenario, noe S with transmission range t S r communicates the sense ata towars the sink through noe the noes i, j, an k with transmission ranges t i r, t j r, an t k r, respectively. Sea Surface Sinks Seabe tr S Figure 1. UAN Scenario. As state earlier, the electromagnetic or raio an optical communications moes can also be use in the unerwater environment; however, they are not suitable ue to ifferent aquatic characteristics. Electromagnetic signals at raio frequencies are preferable for the terrestrial communication because they consume low energy, offer high banwith, small propagation elay, an fairly low signal attenuation. Hence, the Sinks use the raio communication moe to connect the onshore stations. However, raio communication in the aquatic environment is not suitable because of the high attenuation that results in the severe absorption because of the water conuctivity. In result, the raio communication has a limite small transmission range in the aquatic environment. Similarly, the optical communication requires a very precise line of sight between receiver an Sener noe pair as well as the clear visibility. These two constraints are ifficult to meet because of the water current an its turbiity. Hence, the optical communication fails to achieve the high ata rate communication over the long range. Because of these challenges an limitations, the acoustic communication moe is wiely use in the unerwater networks. However, the ata collection an issemination solutions for the terrestrial networks can not irectly be applie in the unerwater network ue to ue to its inherent limitations. There are many routing protocols in the literature that tackle an consier the intrinsic acoustic parameters (e.g., high energy consumption, long propagation elay, large error rate, low banwith
Sensors 2017, 17, 2212 3 of 25 an so forth) to perform routing ecisions. There are many ways the authors have classifie the routing protocols in the literature, however, here we classify those protocols into two main categories, Location or localization-base an epth base routing protocols. In Location or localization base routing protocols, each noe knows its own three-imensional location, the location of the estination noe(s) or sink(s), an sometimes the location of the neighboring noes [10 14]. Each packet contains the location of the estination noe as well. The main objective of these routing protocols is to irectionally forwar the ata packet towars the sink(s) to minimize the propagation elay, reuce energy consumption an utilize minimum banwith by restricting the broacast of multiple copies of the ata packet. Another category of routing protocols consists of the routing schemes that o not use any location information but the epth of the noe [15 20]. Noe s epth information can be estimate locally by the noe itself from the pressure sensor, an it oes not require any aitional control overhea. The packet oes not contain any estination information because it is assume that the packet nees to be forware towars the surface. Aitionally, the packet contains the epth of the relaying noe. In the context of this paper, we focus on the epth base routing schemes. The epth base routing schemes minimize the propagation elay by istributively selecting the next forwarer noe among the neighbors that has the smallest epth than the packet forwarer or Sener noe. All the other noes at the epths between the packet forwarer an the lowest epth noe will simply iscar or suppress the ata packet transmission [15]. This packet suppression at the intermeiate noes minimizes the energy consumption an implemente using the holing time. In the holing time strategy, each noe computes its holing time when it receives the ata packet an the holing time computation simply involves the epth information of the noe(s). Before computation of the holing time, each noe must check whether its epth is smaller than the epth of the ata packet forwaring noe. If a noe has the smaller epth, then it estimates the holing time. Otherwise, it simply iscars the packet. Holing time of the smaller epth noe(s) that the noe(s) with the smallest epth in the neighborhoo of the ata packet sener must have the smallest holing time among all neighbors. After computation of the holing time, it initiates the timer. Once the holing timer expires, the noe verifies whether it has receive aitional copies of the same ata packet or not. If it i not receive any aitional copy, then it further transmits the packet. Otherwise, it suppresses the transmission an removes the copy of the ata packet from its buffer. To ensure that the smaller epth neighbor must forwar the packet, its holing time must be shorter than rest of the neighbors. Along with the epth information, there are some epth base routing schemes that consier not only epth but other metrics to forwar the ata packets or compute the holing time. These metrics not only involves the epth but also the energy of the noe, link quality, epth information of the next hop forwarers an so forth [16 20]. The propagation elay in the epth base routing scheme can be minimize if the noe with lesser epth forwars the ata packet because that noe is closer to Sink. However, this epth base forwaring process may penalize the less epth noes, an all the communication is forware through the same set of noes. In result, the battery power of these noes will rain very quickly an create the voi region. The phenomenon of the voi region is simply escribe as the area in the network where there is no noe available to further the communication towars the estination noe that is Sink. Hence, any noe at the ege of the voi region fails to forwar the ata packets. In orer to avoi the communication through ege noe of the voi region, the schemes in literature consier the epth ifferences between the sener, receiver an one-hop neighbors of receiver noes. It assures that the potential forwarer must have the next hop noe, not the voi region. However, this type of schemes o not avoi the formation of the voi region an o not provie any energy balancing in the network. Therefore, the routing schemes shoul avoi communication through voi region an balance the energy by istributing the ata broacast loa evenly on ifferent noes in the network. The lowest epth noes with small resiual energy must not be penalize. In this manner, we can elay the creation of the voi region. However, if there is alreay a voi region in the network, then the routing scheme must also ivert the communication aroun the voi area.
Sensors 2017, 17, 2212 4 of 25 By keeping all those features uner consieration, we have propose an Energy an Depth variance-base Opportunistic Voi avoiance (EDOVE) scheme in this paper. EDOVE uses the epth ifference information between the ata packet receiving noe (simply we call it receiver noe in this context) an the sener ata, forwaring or relaying noe (call it Sener noe in the context of this paper.) as well as the neighbors of the receiver noe. EDOVE prioritize the receiver noe to become the potential forwaring caniate, if: it has maximum resiual energy among its one hop neighbors. it has one or more than one neighbors with less epth. it has more number of neighbors with epth less than the receiver noe an their normalize epth variance shoul be larger. The above mentione characteristics select a potential forwarer that has multiple noes with lower epth to avoi ata packet collision at the next hop, oes not penalize the low battery power noes, an avois the potential voi region. The rest of the manuscript is organize in the following manner: The literature review that is relate to the propose EDOVE is summarize in Section 2. In Section 3, holing time an working principle of EDOVE is escribe. Simulation analysis of EDOVE is compare to the state of the art scheme in terms of en-to-en elay, energy consumptions, the number of copies of ata packets, an hop count, packet elivery ratio, ea noes, etc., are iscusse in Section 4. Finally, the overall conclusion extracte from the iscussion is presente in Section 5. 2. Previous Works There have been many routing schemes propose to achieve efficient routing in the unerwater acoustic networks[21 23]. In literature, there are many schemes that avoi the uplicate packet forwaring using the backoff timers in the flooing-base unerwater acoustic networks, as in [24]. When a noe receives the ata packet, it initiates the backoff timer an hols the communication of that packet. If it receives a certain number of copies of the same ata packet, which is f loor(2.5), the noe simply iscars the packet. The timer is selecte ranomly between the preefine minimum an maximum backoff uration that is 5 s an 65 s. This scheme oes not use any aitional network parameter in the forwaring ecision. The above work was extene by authors in [25], where they use hop count information, aitional to the backoff, to better ecie which packet shoul be consiere uplicate an iscare. There are many similar works in the literature that involve the backoff timers, however, in this section, we iscuss the epth base opportunistic voi avoiance schemes, which are closely relate to our propose EDOVE scheme. As iscusse in the previous text, the epth information is the prime metric that is use by the routing protocols. It is estimate locally from the reaing of the pressure sensor. Data packets are forware from the higher epth noes to the lower epth towars the surface where sink(s) that are floating on the water surface. In aition to the epth information, several other metrics have been aapte by the routing schemes to efficiently forwar the ata packets to sink(s). Depth base routing schemes are further categorize into two categories; sener initiate (The next hop forwarer of the ata packet is selecte by the Sener noe) [17,19] an the receiver initiate (where the ata packet receiver ecies whether it forwars the packet or not) [15,16,18,20]. In the sener initiate protocols, the ata packet Sener noe selects the next hop relay noe, an the information of the next forwarer is ae in the ata packet. Once the noe receives that ata packet, it easily etermines from the heaer information of the ata packet that noe has to further relay the packet or not. On the other han in the receiver initiate routing protocols, the ata packet receiving noes istributively conten to be the next forwarer. These noes compute their respective holing times, an the potential forwaring caniate(s) has/have very small holing time compare to the other noes. The noe with smallest holing time forwars the ata packet, an when the other noes receive a copy of the same ata packet while their holing time is still on, the simply iscar the
Sensors 2017, 17, 2212 5 of 25 transmission [15]. In this manner, the receiver initiate forwaring schemes suppress the ata packet broacast. Hence, the ata packet broacast suppression epens upon the variance of the holing times compute by the noes. In this section, we merely focus on the receiver initiate epth base routing schemes because the propose EDOVE belongs to that routing category. In [15], authors propose a holing time base packet suppression scheme where packet receiving noes take the forwaring ecision using their epth information, calle Depth Base Routing (DBR). All the noes that are in transmission range of the sener an have smaller epths are caniate forwarers. Rest of the noes that receive ata packet an are place at the larger epth will simply iscar the ata packet. Consier the DBR scenario in Figure 2, where Sener noe S either sens or forwars the ata packet. All noes within the transmission range of S, t S r, e.g., noes 1 to 5, receive the ata message. The receiving noes 2, 3 an 4 with epths 2, 3 an 4, respectively, are the potential forwaring caniates because their epths are less than the epth of S. Accoring to DBR, noe 3 is a preferable forwarer ue to the lowest epth among all the other noes. Conversely, the noes 1 an 5 have higher epth than S. Hence, they will iscar the packet. All potential forwarers in DBR compute their holing time an the smallest epth noe has the very short holing time. DBR computes holing time using the epth ifference between Sener an receiver noes. Hence, the noes with smaller epth an very large epth ifference will always have the short holing time, such as noe 3 in the scenario shown in Figure 2. In other wors, among all the caniate noes, the noe with minimum epth always forwar the packet in the upstream irection. Whenever the same set of seners forwar the ata packet(s), that particular set of noe(s) in the neighborhoo with minimum epth are penalize an ie out earlier. The number of reunant packet transmissions increases with the increase in the network ensity because the elay ifference between their holing times become negligible. Accoringly, excessive energy is consume, an more packets fail to reach the sink noe. 9 t 3 r 2 3 4 6 2 3 4 Seabe 1 S 5 t S r Figure 2. Depth base routing scenario. Authors in [16], propose an Energy Efficient Depth Base Routing (EEDBR). EEDBR computes the holing time using noe s resiual energy an epth information to forwar the packet. Every noe maintains the neighborhoo list where it recors the noe ientifiers (IDs) along with their resiual energy. In EEDBR, the ata sening noe selects the set of potential forwarers with smaller epths from the neighborhoo list an IDs of those potential forwarers are ae in the ata packet. In aition to that, the require packet elivery ratio set by the application is also ae in the ata packet heaer. When a noe receives the ata packet, it first checks whether it is in the potential forwarer set or not. If a noe is not in the potential forwarer set, then it simply iscars the packet. On the other han, if a noe is in the potential forwarer set, it compares its resiual energy with the resiual energy of its neighbors in the neighborhoo list. In case, the noe has maximum resiual energy among its neighbors; it will forwar the packet immeiately. Rest of the potential noes with less resiual energy, compute an start their holing timer an efer the transmission. If any noe
Sensors 2017, 17, 2212 6 of 25 uring its active holing timer receives a copy of the same ata packet, it will compute a ranom number. This ranom number is use to ecie whether to forwar the ata packet or not once the holing timer expires. If the ranom number is less than the packet elivery ratio specifie in the ata packet an the noe receive multiple copies of the ata packet, then it simply iscars the packet. A noe only forwars the ata packet, after its holing timer expires, if any of the following conitions meet; either its compute ranom number is larger than the elivery ratio specifie in the ata packet heaer, or it i not receive any other copy of the same ata packet. EEDBR fails to suppress multiple transmission of the packets an the number of packet transmissions can greatly increase in the ense network scenario. Another epth, hop count, an link quality base routing scheme, calle location-free link state routing (LLSR), has been propose in [17]. The proximity of a noe to the Sink is euce from the hop-count metric an authors assume that this metric avois the communication through the voi region. The beacon issemination phase starte by the Sink with zero hop count value an any noe that receives this beacon message just increments an recors the receive hop count value an then further isseminates the upate hop count in the same beacon message. In aition to the hop count metric, a noe also stores the path quality towars the sink. Subsequently, each noe perioically broacasts its epth, path quality, an hop-count metrics uring the beacon issemination phase. Each noe maintains the 1-hop neighbor table to make the forwaring ecision. The ata packet forwaring noe first checks the lowest hop-count noe neighbor to select it as a potential forwarer. In a tie situation, where two or more than two neighbors have the minimum lowest count, the ecision is taken base on the path quality. If still that tie persists, then the next tie breaker metric is the epth of the noe. The lowest epth noe will get the chance to forwar the ata packet. The packets will be forware through the same set of noes that have lowest hops, goo path quality, an smaller epth. In result, it will penalize the low energy noes an they will ie out quickly. Moreover, the beaconing overhea to maintain the soft-state information is very high an that rastically consumes the network energy. Another receiver initiate voi avoiance routing protocol, calle inherently voi avoiance routing (IVAR), has been propose in [18]. IVAR forwars packets towars Sinks using two routing metrics: epth an hop count of a noe. In orer to maintain the hop count at each noe, all Sinks perioically broacast the beacon message. This beacon message is forware as in [17]. Each ata packet inclues the hop count information. When a ata packet is forware by the noe, then all the receiving noes that have the hop count smaller than the one in the ata packet are eligible caniates. These forwaring noes compute their holing time using their respective epth information. In IVAR, the noes at the ege of a voi region simply iscar the transmission, because their hop count is either equal or greater than the one in the receive ata packet. IVAR avois the voi region an the communication is carrie out through the noes with the small hop count only an it fails to achieve an balance the energy efficiency. The authors in [19] improve the shortcomings of the IVAR an propose a sener initiate opportunistic voi avoiance routing (OVAR). These shortcomings inclue; the uplicate packet transmission that is resulte by the small ifference in the holing times of the caniate forwarer an the hien terminal problem of IVAR. OVAR aapts the same perioic beaconing metho to maintain information of the 1-hop neighbors. This 1-hop ajacency information at each forwarer helps it to select the next caniate forwarers. The caniate noes are selecte in a proximity to avoi the hien noe problem an reuce uplicate packet transmission. In the above-iscusse forwaring protocols that use beacons to upate proximity graph, e.g., LLSR, OVAR an IVAR, neighborhoo information freshness epens on the beacon interval. The lower the beacon interval, the higher the neighborhoo information accuracy an better the forwaring ecision. However, the low beacon interval has severe impact on the battery power. Recently, a receiver initiate epth base routing protocol, name weighting epth an forwaring area ivision DBR (WDFAD-DBR), has been propose that avois the voi hole problem in UWSNs [20]. It achieves energy efficiency in the ense network scenario by suppressing the uplicate ata packets an reucing en-to-en elay by consiering the two-hop istance value in the holing time calculation. Routing in WDFAD-DBR is
Sensors 2017, 17, 2212 7 of 25 performe into two phases: First, the forwaring region is ivie into one primary an two auxiliary forwaring areas to increase the chances of packet elivery. Secon, it computes the weighte sum of the epth ifferences of two hops to minimize the packet uplication. This two hop information prevents the voi hole encounter, however, merely two hop epth ifference oes not eliminate the chance of voi hole encounter. The probability of packet rop an en-to-en elay increases with the increase in the network ensity because the holing time oes not use any noe ensity information. Figure 3 shows the voi region scenario where noe S sens the ata packet an noes 1 5 are within transmission range of noe S, tr S. Noes 2 4 are the forwaring caniates an the noes 1 an 5 simply rop the packet because they have larger epth than S. As noe 3 has no next hop forwarer, hence, it oes not compute the holing time an simply iscars the packet. However, noes 2 an 4 compute their holing times using two hop epth ifference; epth ifference to S an maximum epth ifference to their neighbors with the lower epths, ( 2 an 11 NF), an ( 4 an 7 NF ), respectively. Let, noe 4 has the largest two hop epth ifference, which makes it an appropriate forwarer. It is worth noting that the neighbors of noe 4 are in a close proximity. Therefore, once the noe 4 relays the packet an its neighbors conten to further forwar the packet, it will increase the probability of uplicate packets, collision, an large energy consumption. Similarly, if the resiual energy of noe 4 is very small, then every time S forwars the ata packet through noe 4 an it consumes the energy of noe 4 an its neighbors an eventually will wien the voi hole immeiately. Sea Surface t3 r Voi Region t2 r 11 NF 10 11 NF 10 9 NF 9 Seabe 1 2 2 S 3 4 4 5 NF 6 Note: S is the Sener noe. Noes 2, 3, an 4 are the receivers. 7 6 NF 7 ts r Figure 3. WDFAD-DBR routing scenario. 3 8 NF 8 t4 r Motivation an Contributions Most of the epth base forwaring schemes in the literature just consier epth plus energy of the noe itself to forwar the packet. There are couple of schemes that consier link quality an hop count. However, they nee to exchange a lot of beacons or short messages to get the current link quality an to upate the hop count in the link breakage situation. This beacon control overhea has a rastic impact on a noe s battery an the overall network lifetime. Inspire from the above iscussion, we propose a novel EDOVE for UASNs. EDOVE computes the holing time of caniate forwarers by keeping the following points uner consieration.
Sensors 2017, 17, 2212 8 of 25 1. Avoi using constant weighting factor(s) to compute the holing time because the estimation of the optimal value quite ifficult in the unpreictable an error prone topology network, e.g., α weighting factor was use by WDFAD-DBR. 2. Select the forwarer from the 1-hop neighbors of the ata Sener noe (also known as the ata receiver noes) that has smaller epth relative to the sener noe an has more resiual energy. To achieve this, we use epth ifference between sener an its 1-hop neighbors (the first hop epth ifference parameter) an it is prioritize using the normalize resiual energy of the potential forwarer. The main objective of prioritizing the smaller epth potential forwarer with larger resiual energy is to achieve energy efficiency, energy balancing, an isseminating ata further in the network. In result, it extens the network lifetime an minimizes the ata broacast overhea. 3. Next, choose the forwarer that has larger epth ifference relative to its neighbors, which is the secon hop epth ifference, an has more epth variance among its neighbors. The secon hop epth parameter is scale with the epth proximity variance of the neighbors of the potential forwarer noe. Any potential forwarer that has the large epth proximity variance becomes the potential forwarer an it intens to avoi the packet collision an prevents the voi region when it further forwars the packet. 4. The aaptive weighting factors, base on 2 the resiual energy an 3 the epth variance, with one an two hop epth ifference parameters, significantly scale the holing time of the receivers to avoi multiple ata packet transmissions. In the next section, we briefly iscuss our propose EDOVE scheme an next we iscuss the simulation results, which shows that EDOVE improves energy efficiency, increases packet elivery ratio, an network lifetime. 3. Propose EDOVE Scheme In this section, we present the etaile iscussion of our propose EDOVE scheme. EDOVE avois the voi region by istributively selecting the forwarer with large resiual energy an having multiple neighbors that are largely istribute epth-wise. The propose scheme is base on an compare with the WDFAD-DBR, which avois the voi hole an uses the holing time (T h ) to suppress the uplicate packet transmission. Voi hole avoiance is achieve through prioritizing the forwarers that have one or more than one neighbors with the smaller epth. WDFAD-DBR uses a constant weighting factor, α (0, 1), to prioritize either first hop (sener S an receiver i s epth ifference, i = (z S z i ) an secon hop epth ifferences (receiver an its minimum epth next hop neighbor j, i NF = (z i min(z j )), H. H = α( i ) + (1 α)( NF j ) In aition to that, a maximum acoustic elay between all the receivers, β, is multiplie with H to compute the holing time at receiver i, T hi : T hi = β(t S r H) where β = 2tS r /v soun ρ, ρ (0, tr S ], tr S is the transmission range of S, an v soun is the soun velocity in the aquatic environment. Opposite to WDFAD-DBR, the propose EDOVE computes T hi using the weighte two hop epth ifferences H, the normalize resiual energy of noe i, e i an the next hop epth ifference variance, δ i. Hence, the resultant T hi prioritizes the receiver that has more available energy, large epth ifference to S an receiver with large epth ifference variance to minimize further ata collisions. As a result, energy efficiency is achieve by avoiing the packet collision with voi avoiance an increasing the network lifetime.
Sensors 2017, 17, 2212 9 of 25 3.1. Problem Statement Consier the unerwater network scenario in Figure 3 with sener S, receivers with resiual energy E i within the tr S of S, an receiver with small hop neighbors. The sener S either generates or relays the ata packet an all the receivers simply receive that packet an that packet has to be relaye through any of these receivers. However, these receivers are at the ege of the voi zone an the forwaring algorithm must avoi selecting the relay that fails to further relay the packet ue to the voi zone. Furthermore, these receivers have ifferent resiual energy an the forwaring strategy has to select the receiver with large resiual energy to balance network-wise energy consumption. Selection of one receiver among multiple potential ones is one through the holing time of the receiver i, T hi. When receiver i receives a ata packet, it computes an starts the T hi. While T hi is running an i receives another copy of the same packet, it suppresses the packet transmission. Nonetheless, i forwars the ata packet if it oes not receive any more copies of the ata packet until the T hi expires. However, the computation of T hi must consier the long propagation elay between the receivers an must involve the weighte epth ifferences, the normalize resiual energy of the receiver, an variance of the epth ifferences of the next hop neighbors of i. So, the uplicate packet broacast along with the voi hole can be avoie. However, the impact of this long holing time may result in the long en-to-en elay. Next, we iscuss the propose EDOVE an its holing time computation on how it suppresses the aitional ata packet an avois the voi area in the unerwater network. 3.2. Preliminary Notations EDOVE uses the notations in the holing time computation an the forwaring algorithm, an these notations are briefly escribe below with reference to the EDOVE network scenario shown in Figure 4. Network size N : is the total number of noes N with the acoustic communication capabilities that form the network, is terme as the network size. The set of receiver noes κ S : consists the noes that are within the transmission range of S an have epths less than the epth of S: κ S = {j N z j < z S ist (j,s) t S r } (1) where, ist j,s is eucliean istance between noe j an S, an z j an z S are the epths of noe j an S, respectively. Figure 4 shows that the noes 1 to 5 are the neighbors of S an κ S = {2, 3, 4}. Next hop forwarers of noe i, κi NF : are the noes that lie within noe i s transmission range an having epths less than the epth of i: κ NF i = {j N z j < z i ist (i,j) t i r} (2) where, z j an z i are the epths of next hop noe j an receiver noe i, respectively. The next hop forwarer of receiver 4, κ4 NF = {3, 6, 7, 8}, as shown in Figure 4. However, κ3 NF =. Resiual energy of noe i, E i : is the current available battery level of noe i. Normalize epth variance of noe i, δ i : is the normalize variance of the epth ifferences of all the noes in κi NF. The above iscusse notations are use by the EDOVE to compute holing time an the forwaring algorithm that are briefly escribe below.
Sensors 2017, 17, 2212 10 of 25 Sea Surface t2 r 11 NF 10 11 NF 10 9 NF 9 Seabe 1 t3 r 2 E 1 2 Voi Region E 3 S 3 4 E E2 4 3 4 5 7 7 6 E 6 NF 6 NF ts r 8 NF 8 Note: S is the Sener noe. Noes 2, 3, an 4 are the receivers. E i is the resiual battery level of a noe i. is the epth ifference variance of next hop forwarers. Figure 4. EDOVE network parameters to compute holing time. E 5 E 7 E 8 t4 r 3.3. Estimation of T hi Let, noe S sens or forwars the ata packet an noe i within its tr S receives that ata packet. Then, the noe i first compares its epth to the epth of S. If i has the epth larger than that of S, z i > z S, then it simply iscars the ata packet. On the other han, if noe i s epth is less than the epth S an has one or more than one neighbors with smaller epth (i.e., z i < z S an κi NF 1), then it computes the holing time T hi. However, if noe i has no neighbor at the epth lower than i, then it also iscars the ata packet. The holing time of noe i shoul be very small compare to its neighboring noes if it has the following characteristics: larger resiual energy than its neighbors. larger epth ifference than sener. larger epth ifference to its minimum epth neighbor. many neighbors with large variance in their epth ifferences. EDOVE consiers all the above mentione characteristics to calculate the holing time of i κ S as follows: where, T hi = 2 ts r v Soun (H) (3) ( H = e i 1 ) ( i tr S + 1 max(nf tr i j ) ) δ i, (4) j = 1,..., κ NF i, v Soun 1500 m/s.
Sensors 2017, 17, 2212 11 of 25 Equation (4) consists of two parts. First part prioritizes a noe from the 1-hop neighbors of the ata sener that has smaller epth with larger epth ifference i relative to the Sener noe an large normalize resiual energy, e i, which is explaine later in this section. The main objective of prioritizing the smaller epth potential forwarer with larger resiual energy is to achieve energy efficiency, energy balancing, an isseminating ata further in the network. In result, it extens the network lifetime an minimizes the ata broacast overhea. Next part of Equation (4) consiers the epth ifference between the potential forwarer i relative to its neighbors, max( NF j ), an epth variance among its neighbors δ i. Any ata receiving noe that has the large epth ifference to its lowest epth neighbor an has large epth variance becomes the potential forwarer. This part of the expression intens to avoi the packet collision an prevents the voi region when it further forwars the packet. The factor, e i, in Equation (4) is the normalize resiual energy of i compute using the resiual energy of all the neighbors of i, κ i. e i = E max E i E max E min (5) where, E i = Resiual Energy of noe i, E max = E min, E min = min (E k k κ r ), E max = max (E k k κ r ), an e i [0, 1]. In case, E max = E min, then e i = 1.0. The value of e i will be smaller for the noe i with large resiual energy an that noe will be prioritize to forwar the ata packet. Similarly, the secon factor, δ i, in Equation (4) is the normalize variance of the epth ifferences of the next forwaring noes of i, δ i. δ i = 1 κ NF i 1 κnf i 1 j=1 µ ( ) 2 NF j µ /t i r (6) where, µ = 1 κ NF i κi NF m=1 ( NF m ) The value returne by Equation (6) will be smaller for the receiver i that has the large normalize epth ifference variance, an that noe will be prioritize to act as the forwarer noe. Hence, the propose holing time in EDOVE will prefer the noe that has large resiual energy an has multiple smaller epth noes with large epth variance to further avoi the ata packet collision. Since, computation of the holing time requires resiual battery power an istance information of its one-hop neighbor, therefore, each noe exchanges this information in the neighbor request (NBR_REQ) an neighbor acknowlegment (NBR_ACK) packets, as in [20]. This information is tabulate in the neighbor table (NBR_TABLE). A etaile working principle of our propose EDOVE scheme is illustrate in Algorithm 1. It is very obvious from the Algorithm 1 that once a noe i receives the ata packet, it first checks whether the ata packet is alreay in packet queue, Q, or not. If i i not alreay process the ata packet, then it verifies the timer for that packet. Next, i compares its epth to the one in the ata packet. If i is within the receivers of ata κ i an has more number of neighbors with less epth κi NF, then it
Sensors 2017, 17, 2212 12 of 25 computes the T hi. Otherwise, i just upates the neighborhoo table an rops the ata packet. When i computes the T hi, then it sets this holing time value to the timer (Timer ata = T hi ) an starts. While the ata timer is running an i receives another copy of the packet, then it just iscars the packet an upates the number of copies receive count. Once the Timer ata is triggere, i checks whether it a single copy or multiple copies of the ata packet. The noe i rops the packet if it receives multiple copies. Otherwise, it forwars the ata packet with the upate epth, energy, etc. parameters in the packet. Algorithm 1: Propose EDOVE Scheme. Output:Forwar or Discar ata packet. Input :Noe i receives ata packet containing {ID S, z S, E max, E min, tr S, Contents}. 1 Q= Packet Queue. 2 κ i = List of receivers. 3 κi NF = List of next-hop forwarers of i. 4 Timer ata = Timer for the receive ata packet. 5 if {ata / Q} then 6 A ata in Q 7 if Timer ata is OFF then 8 {t S r, z S, E max, E min } get ata // When noe is the receiver. 9 if i κ S then 10 if κi NF = 0 then 11 Compute e i as in Equation (5) 12 Compute δ i as in Equation (6) 13 Calculate T hi 14 Set Timer ata = T hi 15 Start Timer ata 16 Call 17 en 18 en 19 en 20 Remove ata from Q 21 Upate One-hop NBR_Table 22 Drop ata 23 Exit 24 else 25 Increment the receive ata packet count. 26 Upate One-hop NBR_Table 27 Drop (ata) 28 Exit 29 en 1 Proceure ata,q 2 if ata Q & Count of receive ata > 1 then 3 Remove ata from Q: Q = Q \ ata 4 Exit 5 else 6 Compute E max an E min from NBR_Table 7 Estimate epth z i Upate i, z i, E max, E min an t i r in ata 8 Forwar ata 9 Remove ata from Q: PQ = PQ \ p 10 en 11 return
Sensors 2017, 17, 2212 13 of 25 4. Simulation Analysis The etaile performance comparison of the propose EDOVE scheme in contrast to the conventional WDFAD-DBR scheme through simulations is presente in this section. During simulations, we consiere the three-imensional network area of 1 km length an with an 2 km of epth, (for example, X max = Y max = 1 km an Z max = 2 km). A varying size network of 60, 90,..., up to 180 noes has been consiere in the simulations. All noes are homogeneous in terms of transmission range, which ranges from 600 m to 1000 m. These parameters are sufficient enough to emonstrate the network s spares an enseness. During all simulations, the ata generating noe or source noe is place at the bottom of the ocean an the sink(s) are place at the sea surface. In this scenario, the impact of epth as well as multi-hop forwaring on the epth-base propose solutions can easily be investigate. Rest of the noes are ranomly eploye between source an sink noe(s) in each simulation scenario. These noes serve as the ata forwarers or relaying noes between ata source an the sink(s). This typical scenario explicitly emonstrates, how the ata relaying loa among forwarers is istribute by the forwaring scheme. In first simulation trial for 60 noes, except ata source, all noes are ranomly eploye in the network area. In the next trial for 90 noes, the first 60 noes position is un-altere an only the next 30 noes are ranomly place in the network. There are five sinks eploye over the surface of the network area. It is assume that the whole network is static throughout the simulation uration. Each noe is assigne initial energy E 0 that is equal to E min + ran(e ran ), Where E min = 60 J an E ran = 20 J. We use the simplifie energy consumption moel at each noe. Energy consume by a noe to transmit m bits at istance k meters is presente as: E tx (m, k) = E ele m + E amp m k 2 where E ele is the energy require to sen 1 bit an E amp is the acoustic amplifier energy or acoustic wave attenuation coefficient. Energy consumption at the noe, which receives m bits is represente as: E rx (m) = E ele m. The noe s sening, receiving, an ile energies are 50 W, 158 mw an 58 mw, respectively. The ata, control traffic, an acoustic network parameters use in the simulations are as follows: The heaer an payloa size of ata packet is 88 an 576 bits, respectively. Neighbor request an acknowlegment are 48 bits. The common heaer of 88 bits is use for all packet types in the simulation. Aitionally, the acoustic propagation elay of 1500 m in the aquatic environment an the ata rate of 16 10 3 bits per secon have been consiere in the simulations. In last, at MAC layer, the pure ALOHA is use because of its is not susceptible to elays an oes not use any collision etection an avoiance mechanism. As state earlier that the source noe is eploye at the seabe an it generates ata packets estine for the sink noe(s). The average ata packet generation rate is 0.2 packets per secon or one packet per 5 s. Aitionally, there are some constants that use by the WDFAD-DBR algorithm, e.g., weighting factor α an global parameter τ. The value of α = [0, 1] prioritizes the epth ifferences (between Sener an receiver noe an the receiver itself an its lowest epth neighbor) in the holing time computation of the WDFAD-DBR. To equality prioritize both the epth ifferences, we use α = 0.5. Similarly, the global parameter τ ha been use to guarantee the holing time between two noes shoul be large enough to guarantee packet suppression. About 100 simulations of a single network scenario (a network with the specific number of noes an the transmission range) are performe. The final results are the average of simulation trials an presente with the 95% confience interval in the graphs. Simulation uration for a single soun was 500 s. The performance metrics that are measure through simulations are iscusse in etail in the next section.
Sensors 2017, 17, 2212 14 of 25 4.1. Broacaste Copies of Data Packet: The first performance metric that we investigate here is the number of copies forware in the network for all ata packet transmissions initiate by the source noe. Figures 5 an 6 show the total forware copies of ata packets in the network for ifferent size networks an transmission ranges, respectively. It is observe from Figure 5 that the number of forware copies of ata packets in the network is irectly proportional to the network size for the fixe transmission range. This phenomenon is obvious because in the large network size there may be many uplicate ata packet transmissions. It is because the small network size an fixe transmission range o not cover the whole network, therefore, when the network size increases, more copies of the ata packets are forware within the entire network area. In aition to that, there is a high probability of ifferent noes with similar epth ifference an the epth ifference between their next hop neighbors when network size increases. Hence, their estimate holing times are almost similar that increases the ata packet traffic. In an extreme case, packet suppression can be guarantee by both WDFAD-DBR an EDOVE, where there are multiple noes at similar epths but extreme ens of noe S s transmission range t S r. These noes are not in each other s transmission range an can not suppress the uplicate packet forwaring. This is another reason that most of the noes in the network just forwar the packet. The next interesting tren that has been observe is that the number of forware copies of the ata packets are inversely proportional to the transmission range in the fixe network size scenario. This phenomenon is also apparent ue to the fact that the packet suppression is achieve at many noes in the network when transmission range increases, refer Figure 6. However, it is not possible to guarantee the perfect packet suppression where only one noe within the t S r forwars the ata packet. In result, the chances of packet collision may increase when the transmission range increases because of the large noe ensity within the transmission range. Upon closely observing the simulation results, it is evient that the propose EDOVE forwars less copies of ata packet compare to WDFAD-DBR in any variation of the network parameter scenario. The WDFAD-DBR assigns constant priorities to the epth ifferences, which oes not guarantee the enough holing time ifference between the similar epth ifference noes that are the neighbors of the ata forwarer. Hence, most of the neighbors of the ata forwarer further relay the ata packet. This also increases the packet collision chance an reuces the packet elivery ratio, which is iscusse later in this section. Figure 7 shows the performance of EDOVE in terms of how much less copies of ata are propagate within the network compare to WDFAD-DBR. It shows that EDOVE propagates about 53% less copies of the ata packet than WDFAD-DBR for the largest network size an transmission range. From the above analysis, it is clear that the EDOVE reuces ata broacast storm for varying network sizes an transmission ranges. Next, we analyze the packet elivery performance of the EDOVE an WDFAD-DBR. Total Forware Copies of Data Packets 4500 4000 3500 3000 2500 2000 1500 1000 =600m EDOVE T =1000 r WDFAD DBR T =600m r WDFAD DBR T r WDFAD DBR T =1000m r 500 60 90 120 150 180 Network Size (Noes) Figure 5. Number of forware copies of ata messages in the network for ifferent network sizes.
Sensors 2017, 17, 2212 15 of 25 Total Forware Copies of Data Packets 4500 4000 3500 3000 2500 2000 1500 1000 EDOVE 60 Noes EDOVE 120 Noes EDOVE 180 Noes WDFAD DBR 60 Noes WDFAD DBR 120 Noes WDFAD DBR 180 Noes 500 600 700 800 900 1000 Transmission Range (m) Figure 6. Number of forware copies of ata messages in the network for ifferent transmission ranges. Less Copies of Data Packet Forware (%) 60 40 20 0 1000 900 800 Transmission Range (m) 700 60 600 90 120 150 180 Network Size (Noes) Figure 7. Data packet forwaring loa reuce by EDOVE compare to WDFAD-DBR. 4.2. Packet Delivery Ratio-PDR It is the ratio of successfully receive ata packets by the sinks an the total number of generate ata packets by the source noe. Let, K 1 enotes the set of successfully receive ata packets receive by sink 1 an there are m sinks an n ata packets generate, the PDR is enote as: m K j j=1 PDR = n The main objective of any routing or forwaring scheme is to improve ifferent performance metrics of the network an PDR is one of them. Therefore, we also analyze an iscusse the PDR performance metric of the EDOVE scheme. PDR performance of the EDOVE an WDFAD-DBR for varying network sizes an transmission ranges are shown in Figures 8 an 9, respectively. Figure 8 shows that the PDR of the EDOVE is improve compare to the WDFAD-DBR for ifferent size networks. As we iscusse the phenomenon previously that the number of network-wie forware copies of ata packets is irectly proportional to the network size for the fixe t r. Because of the large ata packet transmissions, the packet collision probability becomes high, an the energy of
Sensors 2017, 17, 2212 16 of 25 the intermeiate relaying noes epletes quickly. As most of intermeiate forwarer noes battery power epletes, they coul not forwar the packets. As EDOVE fairly selects forwarers by using their normalize resiual energy in the holing time computation, it avois penalizing the low battery power noes in the network. Hence, the network using EDOVE survives longer than the network that uses the WDFAD-DBR scheme. The energy consumption an the number of ea noes are also analyze an iscusse later in this section. The ea noes create a voi region an/or a layer is create that completely isconnects the source noe from the sink(s). These events collectively result in a rastic reuction in the PDR. Aitionally, it can also be observe from the figure that as the network size increases, PDR of EDOVE either remains constant or slightly increases, which is not the case of WDFAD-DBR. Next, in the fixe network size an variable transmission range scenario, as the transmission range increases noe ensity increases as well, refer Figure 9. This increases the probability of successfully forwaring the ata packets towar the sink(s) because a large number of reunant transmissions can be suppresse as well as the number of hops between source an sink(s) ecreases. This increase in suppression an ecrease in hop istance, increases the network lifetime, oes not immeiately create any voi region, an more packets are successfully forware towars the sink. On average, EDOVE has 17% more PDR than WDFAD-DBR for any size network an t r = 600 m. Similarly, EDOVE improves about an average of 12% PDR for any t r 700 m contrast to WDFAD-DBR. 0.88 0.86 PDR 0.84 0.82 0.8 0.78 =600m =1000 WDFAD DBR T r =600m WDFAD DBR T r WDFAD DBR T r =1000m 0.76 0.74 0.72 60 90 120 150 180 Network Size (Noes) Figure 8. Average packet elivery ration versus the network sizes. 0.88 0.86 0.84 PDR 0.82 0.8 0.78 EDOVE 60 Noes EDOVE 120 Noes EDOVE 180 Noes WDFAD DBR 60 Noes WDFAD DBR 120 Noes WDFAD DBR 180 Noes 0.76 0.74 0.72 600 700 800 900 1000 Transmission Range (m) Figure 9. Average packet elivery ration versus the transmission range. 4.3. Energy Consumption an Energy Tax In aition to ata broacast overhea an PDR, we also analyze the total energy consumption of the network as the simulation time progresses. The total energy consumption consists of transmission,
Sensors 2017, 17, 2212 17 of 25 reception an ile energy consumption of the network. This total consume energy value is use in the normalize energy consumption that we iscusse previously. This analysis shows the real behavior of the routing scheme propose for the network. Figure 10 epicts the total network energy consumption for the fixe transmission range an ifferent network sizes for the complete simulation course. It is evient from the figure that the total energy consumption of the network is irectly proportional to the network size. However, the propose EDOVE consumes less energy than the WDFAD-DBR. Next, the total energy consumption of the network for the fixe network size an ifferent transmission ranges throughout the simulation course is shown in Figure 11. It shows that there is not that much impact of variation of the t r on the energy consumption of the EDOVE. However, it is noticeable from the same graph that WDFAD-DBR consumes slightly less energy for the t r = 600 m than the t r = 1000 m. In aition to that, it is also apparent from the figure that after 350 s the energy consumption become flat, which shows that there is no more ata forwaring in the network. The main reason behin this occurrence is ue to the fact that there has been a voi layer or most of the relaying noes have been eplete an no more ata coul be forware towars the sink(s). Energy consumption of the WDFAD-DBR saturates earlier than the EDOVE, which shows that most of the relaying noes in WDFAD-DBR run out of the battery power earlier. In the context of this paper, we call this theory as the operation time that is iscusse later in this paper. The operational time is the uration through which ata packet can be forware from the source noe towars the sink(s). Total Energy Consumption (J) 14000 12000 10000 8000 6000 4000 2000 EDOVE: 60 Noes, T r EDOVE 120 Noes, T r EDOVE 180 Noes, T r WDFAD-DBR 60 Noes, T r WDFAD-DBR 120 Noes, T r WDFAD-DBR 180 Noes, T r 0 5 50 95 140 185 230 275 320 365 410 455 500 Simulation Time (s) Figure 10. Total network energy consumption uring the simulation uration for fixe transmission range of 800 m an varying number of noes in the network. 9000 8000 Total Energy Consumption (J) 7000 6000 5000 4000 3000 2000 1000 0 EDOVE: T r =600m, 120 Noes, 120 Noes =1000m, 120 Noes WDFAD-DBR: T r =600m, 120 Noes WDFAD-DBR: T r, 120 Noes WDFAD-DBR: T r =1000m, 120 Noes 5 50 95 140 185 230 275 320 365 410 455 500 Simulation Time (s) Figure 11. Total network energy consumption uring simulation uration for the fixe number of noes (120 noes) an ifferent transmission ranges in the network.
Sensors 2017, 17, 2212 18 of 25 In aition to the network energy consumption analysis, we also compare the propose EDOVE with WDFAD-DBR in terms of the energy tax, E Tax. WDFAD-DBR computes energy tax as the average energy consume per successfully receive ata packet by the sink(s) an per network noe. E Tax = E Total P Success Noes (7) where, P Success is the total number of successfully receive packets at sink noe(s). In case of multiple sinks, P Success is the istinct ata packets that are receive by multiple noes. It is believe the WDFAD-DBR authors that energy tax parameter is more meaningful than total energy consumption. Hence, we compare the performance of EDOVE with the WDFAD-DBR in terms of energy tax for varying network sizes an transmission ranges, which is iscusse below. The energy tax analysis of the EDOVE for ifferent network sizes an transmission ranges are shown Figures 12 an 13, respectively. It shows that the energy tax of the EDOVE is far less than the WDFAD-DBR for ifferent network sizes. The main reason for this behavior is that WDFAD-DBR consumes more energy an has very less elivery ratio. In result, it has very high energy tax compare to the EDOVE scheme. However, it is worth noting that the energy tax tren for both the schemes slightly increases as the network size increases. It is only because the volume of energy consume ue to large copies of ata are isseminate is very high an number of generate ata packets by the source noe are same for any network size. The simulations also prove that the transmission an reception of these ata packets have large impact on the overall energy consumption of the network. 96 94 92 Energy Tax 90 88 86 84 =600m =1000 WDFAD-DBR T r =600m WDFAD-DBR T r WDFAD-DBR T r =1000m 82 80 78 60 90 120 150 180 Network Size (Noes) Figure 12. Energy Tax of the network versus the network sizes. 92 Energy Tax 90 88 86 84 82 80 EDOVE 60 Noes EDOVE 120 Noes EDOVE 180 Noes WDFAD-DBR 60 Noes WDFAD-DBR 120 Noes WDFAD-DBR 180 Noes 78 76 74 72 600 700 800 900 1000 Transmission Range (m) Figure 13. Energy Tax of the network versus the transmission range.
Sensors 2017, 17, 2212 19 of 25 4.4. Number of Dea Noes A noe is consiere ea when its battery is eplete completely an it coul not participate in the ata forwaring process. Figure 14 shows the average number of ea noes for ifferent network sizes an fixe transmission ranges. It inicates that the number of ea noes increases as the network size increases an that is obvious because of a large number of noes participating in the forwaring process. In all the network scenarios with ifferent network sizes, the number of ea noes in the EDOVE is less than the WDFAD-DBR. However, at some cases e.g., network of 180 noes, the number of ea noes in EDOVE scheme are larger than the conventional WDFAD-DBR. This is because of the EDOVE s fair istribution of ata forwaring loa on most of the noes that enable these noes to be active for a longer time an resulte in the large PDR, refer iscussion relate to PDR in the previous subsection. 180 160 Average No. of Dea Noes 140 120 100 80 60 =600m =1000 WDFAD DBR T r =600m WDFAD DBR T r WDFAD DBR T r =1000m 40 60 90 120 150 180 Network Size (Noes) Figure 14. Number of ea noes versus ifferent network sizes. The next etaile analysis relate to the ea noes is shown in Figure 15 that shows the total number of ea noes over the course of the simulation. In the case of WDFAD-DBR, the noes start ying out immeiately after 120 s of the simulation time. On the other han, the noes start ying or epleting their batteries between 180 s to 230 s when EDOVE scheme is execute on the same network scenario because it fairly istributes the ata forwaring loa using their resiual energy an the istribution of their secon hop potential forwarers. It shows that the network using EDOVE survives for a longer uration than the WDFAD-DBR an forwars more ata packets towars the sink(s). 120 2 1.5 100 1 Number of Dea Noes 80 0.5 0 95 140 185 230 120 210 60 225 Noes = 120 EDOVE: T 40 r =600m EDOVE T r EDOVE T =1000m r 20 WDFAD DBR: T r =600m WDFAD DBR: T r WDFAD DBR: T =1000m r 0 5 50 95 140 185 230 275 320 365 410 455 500 Simulation Time (s) Figure 15. Number of ea noes uring simulation uration for the fixe transmission range an ifferent network sizes. Furthermore, we simulate the average number of ea noes for the varying transmission ranges an ifferent network sizes with a fixe transmission range of 800 m over the course of simulation,
Sensors 2017, 17, 2212 20 of 25 as shown in Figures 16 an 17, respectively. Figure 16 shows that the average number of the ea noes increases as the network size increases, which is only because the ata packet traffic rastically increases with the increase in the network size, refer Figure 5. However, it can easily be observe from the figure that EDOVE has slightly less number of ea noes than the WDFAD-DBR. The number of ea noes is an important factor. However, the main factor is the time instance when these noes ie out uring simulation. If they ie out earlier uring the simulation, then they can not properly participate in the ata packet forwaring process. The number of ea noes are analyze for ifferent network sizes an fixe t r = 800 m uring the simulation course are shown in Figure 17. Due to high ata packet traffic for the large network size, it can easily be seen from the figure that noes start ying earlier uring the simulation. For the network of 180, 120 an 60 noes, the noes start ying at 80 s, 125 s an 190 s, respectively, uring the simulation course for WDFAD-DBR. On the other han in the case of EDOVE, the noes start to ie out at the time instance of 185 s, 190 s an 220 s, for the network of size 180, 120 an 60, respectively. From the results, we can easily infer that the network using the EDOVE have slowe or elaye battery epletion rate. 180 Average No. of Dea Noes 160 140 120 100 80 EDOVE 60 Noes EDOVE 120 Noes EDOVE 180 Noes WDFAD DBR 60 Noes WDFAD DBR 120 Noes WDFAD DBR 180 Noes 60 40 600 700 800 900 1000 Transmission Range (m) Figure 16. Number of ea noes versus ifferent transmission ranges. 180 2 160 1 140 Number of Dea Noes 120 100 80 60 40 20 0 50 95 140 185 230 275 80 125 190 220 EDOVE: 60 Noes EDOVE 120 Noes DOVE 180 Noes WDFAD DBR 60 Noes, WDFAD DBR 120 Noes WDFAD DBR 180 Noes T r = 800m 0 5 50 95 140 185 230 275 320 365 410 455 500 Simulation Time (s) Figure 17. Number of ea noes uring simulation uration for the fixe network size an ifferent transmission ranges. 4.5. Average Operational Time The operational time is the time instance uring the simulation when any of the sink(s) receive the last ata message from the source noe. Figures 18 an 19 show the operation time comparison of the EDOVE an WDFAD-DBR for varying network sizes an transmission ranges, respectively. From the previous iscussion, it is clear that the noes in WDFAD-DBR eplete their energy quickly because it oes not consier any energy information in the forwaring ecision. Hence, it fails to maintain energy fairness an potentially creates or extens the voi area or voi layer. In this case,
Sensors 2017, 17, 2212 21 of 25 no further communication can be carrie out towars the sink an the time instance when the very last ata message is receive by the Sink is recore an compare. Figure 18 shows that the average operational time of EDOVE is larger than the WDFAD-DBR. It is also evient from the figure that EDOVE has almost constant operational time for any size network. On the contrary, WDFAD-DBR s operational time ecreases with the increase in the network size. It is only because the WDFAD-DBR generates more ata traffic an epletes energy of most of the noes. Hence, most of the noes near or between the source an sink noe eplete their energy an fail to forwar more ata towar the sink. 360 350 Average Operational Time (s) 340 330 320 EDOVE T =600m r =1000 310 WDFAD DBR T r =600m WDFAD DBR T r WDFAD DBR T =1000m r 300 60 90 120 150 180 Network Size (Noes) Figure 18. Average network operational time versus the network sizes. 360 350 Average Operational Time (s) 340 330 320 310 EDOVE 60 Noes EDOVE 120 Noes EDOVE 180 Noes 300 WDFAD DBR 60 Noes WDFAD DBR 120 Noes WDFAD DBR 180 Noes 290 600 700 800 900 1000 Transmission Range (m) Figure 19. Average network operational time versus the transmission range. Similarly, we recore the average operational time for the varying transmission ranges an show them in Figure 19. It is evient from the figure that operational time of the propose scheme slightly increases as the transmission range increases. However, no marginal variation in the operational time for ifferent network sizes was observe for the EDOVE. On the other han, a significant change in the operational time of WDFAD-DBR was observe. It emonstrate the large operational time for the smaller The reason of the large operation time for smaller network size is that it generates less ata packet broacast. However, the simulation results emonstrate that overall operational time of the network using EDOVE is larger than by using the WDFAD-DBR protocol. The operational time of the EDVOE is almost 40 s more than of the WDFAD-DBR. The overall improvement of the propose EDOVE in contrast to the WEDFAD-DBR is shown in Figure 20.
Sensors 2017, 17, 2212 22 of 25 Increase Operational Time (s) 50 40 30 20 10 0 1000 120 900 90 800 700 60 600 Transmission Range (m) 150 180 Network Size (Noes) Figure 20. Overall increase network lifetime by the propose EDOVE. It has been evient from all the previous results an iscussion that EDOVE has high PDR, less ata broacast, consumes less energy, fewer number of ea noes, an has large operational time. Next, we investigate the average number of hops the ata packet is forware from the source to sink noe(s). Figures 21 an 22 show the average number of hops for ifferent network sizes an transmission ranges, respectively. These figures show that EDOVE has a slightly larger number of hops than the WDFAD-DBR. The reason for this slightly larger number of hops in EDOVE is that it focuses on the reuction of network s energy consumption an increase network lifetime. This is achieve by EDOVE because it uses noe s resiual energy, the variance of the neighboring noes epths an epth ifferences between the sener, receiver, an receiver s next hop forwarers. On the contrary, the WDFAD-DBR just consiers the epth ifference to select the next ata packet forwarer, which reuces the number of hops. This is the reason that WDFAD-DBR has less En-to-En elay compare to EDOVE, refer Figures 23 an 24 for ifferent network size an transmission range. 5.5 Average Hops 5 4.5 4 EDOVE T =600m r EDOVE T r EDOVE T =1000 r WDFAD DBR T =600m r WDFAD DBR T r WDFAD DBR T =1000m r 3.5 3 60 90 120 150 180 Network Size (Noes) Figure 21. Average No. of Hops ata messages traverse versus the network sizes.
Sensors 2017, 17, 2212 23 of 25 Average Hops 6 5.5 5 4.5 4 EDOVE 60 Noes EDOVE 120 Noes EDOVE 180 Noes WDFAD DBR 60 Noes WDFAD DBR 120 Noes WDFAD DBR 180 Noes 3.5 3 600 700 800 900 1000 Transmission Range (m) Figure 22. Average No. of Hops ata messages traverse versus the transmission range. Average En to En Delay (s) 3.6 3.4 3.2 3 2.8 2.6 =600m EDOVE T =1000 r WDFAD DBR T r =600m WDFAD DBR T r WDFAD DBR T r =1000m 2.4 2.2 60 90 120 150 180 Network Size (Noes) Figure 23. En-to-En elay between Source noe that generate ata message an Sink noe versus the network sizes. 0.9 0.85 Average En to En Delay (s) 0.8 0.75 0.7 0.65 EDOVE 60 Noes 0.6 EDOVE 120 Noes EDOVE 180 Noes 0.55 WDFAD DBR 60 Noes WDFAD DBR 120 Noes WDFAD DBR 180 Noes 0.5 600 700 800 900 1000 Transmission Range (m) Figure 24. En-to-En elay between Source noe that generate ata message an Sink noe versus the transmission range. Comprehensively, the propose EDOVE scheme improves the PDR, reuces the ata packet broacast, conserves more energy which in turn reuces the number of ea noes, an has more operational time than the WDFAD-DBR. However, it has slightly longer en-to-en hop istance an slightly larger elay, which is the traeoff of the propose scheme or may be any scheme propose for the UASNs.
Sensors 2017, 17, 2212 24 of 25 5. Conclusions An energy an epth variance base opportunistic voi avoiance (EDOVE) scheme for unerwater acoustic sensor networks has been propose in this paper. EDOVE uses the epth ifference information between ata packet sener to its one-hop neighbor an one-hop neighbor to its forwarers (that are two-hop neighbors of the Sener noe). In aition to that, it prioritizes the epth ifferences using the normalize resiual energy of the forwarer an the normalize epth variance of the two-hop forwarers to compute the holing time. In this manner, it offers energy fairness, istributes ata forwaring loa over the fairly istribute ensity noes an increases the network lifetime. Simulation results show that EDOVE propagates about 53% less copies of the ata packets, increases more than 40 (s) of network life time, reuces energy epletion of the noes than the state of the art WDFAD-DBR. On the other han, EDOVE has slightly longer elay because it focuses on the energy fairness an loa balancing rather than the greey epth base forwaring. This makes EDOVE to be a more suitable technique for UASNs. Acknowlegments: This work was supporte by Institute for Information & communications Technology Promotion(IITP) grant fune by the Korea government(msit) (No. 2014-0-00065, Resilient Cyber-Physical Systems Research). Author Contributions: Safar Hussan Bouk an Sye Hassan Ahme performe the simulations an writeup; Kyung-Joon Park an Yongsoon Eun i the write-up, revision, an overall organization of the manuscript. Conflicts of Interest: The authors eclare no conflict of interest. References 1. Garcia, J.E. Accurate positioning for unerwater acoustic networks. In Proceeings of the 2005-Europe Oceans, Brest, France, 20 23 June 2005. 2. Sozer, E.M.; Stojanovic, M.; Proakis, J.G. Unerwater acoustic networks. IEEE J. Ocean. Eng. 2000, 25, 72 83. 3. Stojanovic, M.; Preisig, J. Unerwater acoustic communication channels: Propagation moels an statistical characterization. IEEE Commun. Mag. 2009, 47, 84 89. 4. Che, X.; Wells, I.; Dickers, G.; Kear, P.; Gong, X. Re-evaluation of RF electromagnetic communication in unerwater sensor networks. IEEE Commun. Mag. 2010, 48, 143 151. 5. Cella, U.M.; Johnstone, R.; Shuley, N. Electromagnetic wave wireless communication in shallow water coastal environment: Theoretical analysis an experimental results. In Proceeings of the Fourth ACM International Workshop on Unerwater Networks, Berkeley, CA, USA, 3 November 2009. 6. Kaushal, H.; Kaoum, G. Unerwater optical wireless communication. IEEE Access 2016, 4, 1518 1547. 7. Hanson, F.; Raic, S. High banwith unerwater optical communication. Appl. Opt. 2008, 47, 277 283. 8. Heiemann, J.; Ye, W.; Wills, J.; Sye, A.; Li, Y. Research challenges an applications for unerwater sensor networking. In Proceeings of the IEEE Wireless Communications an Networking Conference, Las Vegas, NV, USA, 3 6 April 2006. 9. Akyiliz, I.F.; Pompili, D.; Meloia, T. Unerwater acoustic sensor networks: Research challenges. A Hoc Netw. 2005, 3, 257 279. 10. Xie, P.; Cui, J.-H.; Lao, L. VBF: Vector-base forwaring protocol for unerwater sensor networks. In Proceeings of the 5th International IFIP-TC6 conference on Networking Technologies, Services, an Protocols, Performance of Computer an Communication Networks, Mobile an Wireless Communications Systems, Coimbra, Portugal, 15 19 May 2006. 11. Nicolaou, N.; See, A.; Xie, P.; Cui, J.H.; Maggiorini, D. Improving the robustness of location-base routing for unerwater sensor networks. In Proceeings of the 2007-Europe Oceans, Abereen, UK, 18 21 June 2007. 12. Hwang, D.; Kim, D. DFR: Directional flooing-base routing protocol for unerwater sensor networks. In Proceeings of the OCEANS 2008, Quebec City, QC, Canaa, 15 18 September 2008. 13. Ali, T.; Jung, L.T.; Ameer, S. Flooing control by using Angle Base Cone for UWSNs. In Proceeings of the 2012 International Symposium on Telecommunication Technologies, Kuala Lumpur, Malaysia, 26 28 November 2012.
Sensors 2017, 17, 2212 25 of 25 14. Yu, H.; Yao, N.; Liu, J. An aaptive routing protocol in unerwater sparse acoustic sensor networks. A Hoc Netw. 2015, 34, 121 143. 15. Yan, H.; Shi, Z.J.; Cui, J.-H. DBR: Depth-base routing for unerwater sensor networks. In Proceeings of the 7th International IFIP-TC6 Networking Conference on A Hoc an Sensor Networks, Wireless Networks, Next Generation Internet, Singapore, 5 9 May 2008. 16. Wahi, A.; Lee, S.; Jeong, H.-J.; Kim, D. EEDBR: Energy-efficient epth-base routing protocol for unerwater wireless sensor networks. In Avance Computer Science an Information Technology; Springer-Verlag: Berlin/Heielberg, Germany, 2011; pp. 223 234. 17. Barbeau, M.; Blouin, S.; Cervera, G.; Garcia-Alfaro, J.; Kranakis, E. Location-free link state routing for unerwater acoustic sensor networks. In Proceeings of the 28th IEEE Canaian Conference on Electrical an Computer Engineering (CCECE), Halifax, NS, Canaa, 3 6 May 2015. 18. Ghoreyshi, S.M.; Shahrabi, A.; Boutaleb, T. An inherently voi avoiance routing protocol for unerwater sensor networks. In Proceeings of the Twelfth IEEE International Symposium on Wireless Communication Systems, Brussels, Belgium, 25 28 August 2015. 19. Ghoreyshi, S.M.; Shahrabi, A.; Boutaleb, T. A novel cooperative opportunistic routing scheme for unerwater sensor networks. Sensors 2016, 16, 297. 20. Yu, H.; Yao, N.; Wang, T.; Li, G.; Gao, Z.; Tan, G. WDFAD-DBR: Weighting epth an forwaring area ivision DBR routing protocol for uasns. A Hoc Netw. 2016, 37, 256 282. 21. Han, G.; Jiang, J.; Bao, N.; Wan, L.; Guizani, M. Routing protocols for unerwater wireless sensor networks. IEEE Commun. Mag. 2015, 53, 72 78. 22. Li, N.; Martínez, J.-F.; Meneses Chaus, J.M.; Eckert, M. A survey on unerwater acoustic sensor network routing protocols. Sensors 2016, 16, 414. 23. Ghoreyshi, S.M.; Shahrabi, A.; Boutaleb, T. Voi-hanling techniques for routing protocols in unerwater sensor networks: Survey an challenges. IEEE Commun. Surv. Tutor. 2017, 19, 800 827. 24. Otnes, R.; Haavik, S. Duplicate reuction with aaptive backoff for a flooing-base unerwater network protocol. In Proceeings of the 2013 MTS/IEEE OCEANS, Bergen, Norway, 10 14 June 2013. 25. Komulainen, A.; Nilsson, J. Capacity improvements for reuce flooing using istance to sink information in unerwater networks. In Proceeings of the 2014 Unerwater Communications an Networking (UComms), Sestri Levante, Italy, 3 5 September 2014. c 2017 by the authors. Licensee MDPI, Basel, Switzerlan. This article is an open access article istribute uner the terms an conitions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).