Impacts of Deployment Strategies on Localization ... - Semantic Scholar

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 3, MARCH 2015

1725

Impacts of Deployment Strategies on Localization Performance in Underwater Acoustic Sensor Networks Guangjie Han, Member, IEEE, Chenyu Zhang, Lei Shu, Member, IEEE, and Joel J. P. C. Rodrigues, Senior Member, IEEE Abstract—When setting up an underwater acoustic sensor network (UASN), node deployment is the ﬁrst and foremost task, upon which many fundamental network services, such as network topology control, routing, and boundary detection, will be built. While node deployment in 2-D terrestrial wireless sensor networks has been extensively studied, little attention has been received by their 3-D counterparts. This paper aims at analyzing the impacts of node deployment strategies on localization performances in a 3-D environment. More speciﬁcally, the simulations conducted in this paper reveal that the regular tetrahedron deployment scheme outperforms the random deployment scheme and the cube deployment scheme in terms of reducing localization error and increasing localization ratio while maintaining the average number of neighboring anchor nodes and network connectivity. Given the fact that random deployment is the primary choice for most of practical applications to date, our results are expected to shed some light on the design of UASNs in the near future. Index Terms—Localization performance, network connectivity, node deployment, underwater acoustic sensor networks (UASNs).

I. I NTRODUCTION

R

ECENTLY, applications of sensor networks in aqueous environments, such as resource exploration, target

Manuscript received November 16, 2013; revised February 20, 2014, May 23, 2014, and July 20, 2014; accepted September 16, 2014. Date of publication October 13, 2014; date of current version February 6, 2015. This work was supported in part by the Qing Lan Project, in part by the Natural Science Foundation of Jiangsu Province of China under Grant BK20131137, in part by the Applied Basic Research Program of the Nantong Science and Technology Bureau under Grant BK2013032, in part by the Educational Commission of Guangdong Province China Project under Grant 2013KJCX0131, and in part by the 2013 Special Fund of Guangdong Higher School Talent Recruitment. The work of J. J. P. C. Rodrigues was supported in part by the Instituto de Telecomunicaes, in part by the Next Generation Networks and Applications Group (NetGNA), Covilhã Delegation, and in part by National Funding from the FCT-Fundação para a Ciência e a Tecnologia through Project Pest-OE/EEI/LA0008/2013. G. Han and C. Zhang are with the Department of Communication and Information Systems, Hohai University, Changzhou 213022, China (e-mail: [email protected]; [email protected]). L. Shu is with the Guangdong Provincial Key Laboratory of Petrochemical Equipment Fault Diagnosis, Guangdong University of Petrochemical Technology, Maoming 525000, China (e-mail: lei.shu@ ieee.org). J. J. P. C. Rodrigues is with the Instituto de Telecomunicações, University of Beira Interior, 6201-001 Covilhã, Portugal, and also with Saint Petersburg State University of Information Technologies, Mechanics and Optics, Saint Petersburg 197101, Russia (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIE.2014.2362731

Fig. 1.

Three-dimensional UASN architecture.

tracking, and pollution monitoring [1], [2], have attracted rapidly growing attention. Underwater acoustic sensor networks (UASNs) are wireless sensor networks (WSNs) specially designed in aqueous environments for underwater applications, whose implementations and operations are solely based on acoustic measurements and communications. Depending on the application requirements, different kinds of sensor nodes can be deployed in UASNs, e.g., surface sinks meant for data collection and GPS signal acquisition, underwater nodes equipped with floating buoys (which can be inflated by pumps to adjust their depths to ensure full coverage), automatic mobile nodes such as autonomous underwater vehicles, unmanned underwater vehicles, and low-power gliders. In a UASN, events are first detected by sensor nodes locally, and the corresponding information is then forwarded to surface sinks by multihops or automatic mobile nodes via acoustic communications for further processing, as shown in Fig. 1. UASNs are deployed in a 3-D environment, which inevitably brings on new challenges, such as long transmission delay, node mobility caused by water currents [3], [4], etc. Here, we remark that the propagation speed of an acoustic wave is slower than that of a radio wave by five orders of magnitude [5]; moreover, compared with a radio channel, an acoustic channel has higher power loss, more limited bandwidth, and higher bit error rate. These negative factors render it difficult to translate ideas and techniques in the vast literature of 2-D terrestrial WSNs [6], [7] to UASNs. To circumvent these issues, an effective node deployment strategy in UASNs bears even

0278-0046 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

1726


greater importance to ensure network connectivity, seamless coverage, and localization accuracy in UASNs. On the other hand, for UASNs, most of the existing deployment algorithms are designed for 2-D monitored regions, mainly focusing on minimizing the number of sensors to be deployed and energy consumption to prolong network lifetime, ensuring network coverage and connectivity or achieving successful data delivery [8]–[10]. When setting up an UASN, node deployment is the first and foremost task, since it provides fundamental support for many network services, such as network topology control, routing, and boundary detection [11]. In particular, since underwater and bottom nodes transmit the sensing information to surface sinks in a multihop manner, a good deployment strategy will promise network connectivity and provide a stable network topology for the fulfillment of subsequent monitoring tasks [12]. Prominent examples include large-scale short-term and distributed data acquisition networks meant for time-critical aquatic applications [13], [14], where designing a suitable deployment strategy is a basic and must-do task to provide support for network topology control, data collection, etc. Many network applications require the sensor nodes to know their locations with a high degree of precision since such information is indispensable to geographical location routing protocols, which prove to be more efficient than pure flooding in UWSNs [15]. Localization is a process of finding such location information of the sensor nodes in a given coordinate system. To localize a UASN in the global coordinate system, some special nodes should be aware of their positions in advance either from an external GPS or its own memory (initialized during manual placement); such nodes are called anchor nodes (or beacon nodes). Other sensor nodes, which are usually called ordinary nodes (or unknown nodes), calculate their positions by using some localization algorithms. This paper investigates the impacts of different deployment schemes on the localization performances, such as localization error, localization ratio, average number of neighboring anchor nodes, and network connectivity, in 3-D UASNs. To the best of our knowledge, this is the first work that compares the localization performances of random deployment, cube deployment, and regular tetrahedron deployment schemes in 3-D UASNs. The simulations conducted in this paper reveal that the regular tetrahedron deployment scheme outperforms the random deployment scheme and the cube deployment scheme in terms of reducing localization error and increasing localization ratio while maintaining the average number of neighboring anchor nodes and network connectivity. The primary choice for most of practical applications to date is the random deployment, which, despite its simplicity, may cause a variety of issues such as network partition, nonuniform network coverage, etc. In this regard, our results are expected to shed some light on the design of UASNs in the near future. The remainder of this paper is organized as follows. Section II summarizes the related work. Section III describes network model and methodologies. Section IV presents a detailed analysis of simulation results. Section V makes conclusions and discusses future research issues of deployment algorithms in UASNs.

II. R ELATED W ORK Extensive research has focused on node deployment algorithms in UASNs, the majority of which aim at achieving high network connectivity and coverage, minimizing the number of sensor nodes and their energy consumption, or improving data delivery ratio. Alsalih et al. proposed a 2-D node deployment algorithm to find the optimal placement of data collectors and the optimal multihop routing path for delivering data from underwater sensors to onshore data collectors in [9], where the problem has been formulated as an integer linear program. Nie et al. [10] proposed an approach to choose a path to one of the available gateways in light of optimized network performance (minimal delay, minimal energy consumption, and the least packet loss rate). They proposed a scheme to determine surface gateway position based on genetic algorithms for UASNs, which can achieve minimal delay and balance energy consumption throughout simultaneous transmissions. A multipath virtual sink architecture for UASNs was presented in [16] to overcome adverse link conditions. Virtual sinks are deployed at the vertices of the monitored surface. The network dynamically selects the shortest paths to deliver data over multiple routes to increase the probability of successful delivery to avoid retransmissions. The algorithm ensures that data delivery continues to function even when some part of the network is temporarily nonoperational. Alam et al. [17] investigated the maximal coverage and connectivity issues in 3-D networks with the least number of sensor nodes to be deployed in the monitored space. The authors defined a metric called volumetric quotient, which is a metric of the quality of the competing space-filling polyhedrons. They compared volumetric quotients of truncated octahedron cells, hexagonal prism cells, rhombic dodecahedron cells, and cube cells. The results indicated that the use of the Voronoi tessellation to create truncated octahedral cells achieves the best results. In a separate study, Alam et al. [18] assumed that nodes are uniformly and densely deployed in a 3-D space. They divided the 3-D space into cells, in each of which, only one node is active at any given time to minimize the number of active nodes while maintaining the full coverage and connectivity. Results showed that the number of active nodes can be minimized if the shape of each cell is a truncated octahedron and the sensing range is at least 0.52325 times the transmission range. Thus, the truncated octahedron model has the highest network lifetime. In [19], Akkaya et al. proposed a distributed node deployment strategy that can increase the network coverage. In their scheme, sensors were initially deployed at a 2-D bottom region of the ocean, where neighboring nodes compute a certain depth that can minimize the sensing overlaps among themselves. Through a vertex coloring formulation, the so-called redundancy is then determined by one of the neighboring nodes, which will act as the “leader.” The process of depth adjustment continues until there is no room to improve the coverage for a sensor. Although there are many node deployment mechanisms developed for UASNs, the majority of them only focus on how to

HAN et al.: IMPACTS OF DEPLOYMENT STRATEGIES ON LOCALIZATION PERFORMANCE

1727

deploy sensors on the surface or at the bottom of the monitored space, which is a de facto 2-D deployment problem. A notable exception is [3], where three node deployment schemes in a 3-D environment, i.e., 3-D random, bottom random, and bottom grid (see [3, Sec. 5]). The 3-D random scheme in [3], which is the same as the one we studied in this paper, is however only assumed to study the trajectory of a sinking node in a 3-D network. Moreover, the bottom-random and bottom-grid schemes, which deploy sensor nodes on the bottom of the ocean, are essentially 2-D deployment schemes. This paper aims at analyzing the impact of three deployment schemes on localization performances in 3-D UASNs, more specifically, the random deployment scheme, the cube deployment scheme, and the regular tetrahedron deployment scheme, where sensor nodes may be deployed at different depths depending on the deployment strategies. For the three aforementioned schemes, we will examine parameters such as localization error, localization ratio, average number of neighboring anchor nodes, and network connectivity. The conducted simulations show that the regular tetrahedron deployment scheme can achieve higher localization accuracy while maintaining good localization ratio, the average number of sensor nodes, and network connectivity. Detailed performance comparisons and analyses are given in the following. III. N ETWORK M ODEL AND M ETHODOLOGIES To monitor the region that cannot be adequately monitored by only using 2-D nodes, 3-D UASNs are usual alternatives, in which underwater nodes floating at different depths are expected to collaboratively monitor the given region. One promising method is to anchor winch-based underwater nodes at the bottom of the ocean at first, as shown in Fig. 1. In this way, underwater nodes will not drift with water currents. Each underwater node is equipped with a floating buoy that can be inflated by a pump to adjust its depth; specifically, each underwater node, with the help of an electronically controlled engine (installed on the node itself), adjusts its depth by controlling the length of the wire that connects the underwater node to the anchor [3]. In this section, we will illustrate in great detail our network model, where the random deployment scheme, the cube deployment scheme, and the regular tetrahedron deployment scheme will be employed. In the random deployment scheme, anchor nodes are deployed randomly in a 3-D monitored region, as shown in Fig. 2(c). As shown in Fig. 2(a) and (b), in the cube deployment scheme, anchor nodes are deployed at the vertices of the prepositioned space-filling cubes, whereas in the regular tetrahedron deployment scheme, anchor nodes are deployed at the vertices of some prepositioned regular tetrahedrons, as shown in Fig. 2(d). Ordinary nodes in all the three deployment schemes above are deployed randomly in the 3-D monitored space. Note that among all kinds of regular polyhedrons, the cube is the only space-filling regular polyhedron (to fill a volume without any overlaps or gaps) [17]. Here, we remark that, if no prior knowledge of the monitored region is available or deterministic deployment of sensors is very risky or infeasible, the random deployment often becomes the only option.

Fig. 2. Three-dimensional network deployment. (a) Cube depolyment. (b) One cube unit. (c) Random depolyment. (d) One regular tetrahedron unit.

In our model, the following three assumptions are made. 1) Each sensor node has a omnidirectional antenna with the communication range of R. 2) Only if the Euclidean distance between two sensor nodes is not larger than R, they are able to communicate with each other. 3) The deployment region is a 3-D static underwater environment, and the movement of sensor nodes is within the acceptable range, which can be ignored. The transmission delay and transmission loss (TL) of the signals from our underwater acoustic channel are described as follows. • Transmission delay: The propagation speed in the seawater depends on a number of parameters, such as temperature, salinity, and depth; indeed, the increase in any of these three parameters will result in that of the propagation speed. The propagation speed c in m/s can be calculated [5] as follows: c = 1449 + 4.6T + 0.055T 2 + 0.003T 3 + (1.39 − 0.012T )(S − 35) + 0.017D

(1)

where T represents temperature in degree Celsius, S represents salinity in parts per thousand, and D represents depth in meters. Furthermore, the propagation delay t can be calculated as follows: t=

d c

(2)

where d is the distance between two nodes in meters, and c is the sound of speed in meters per second.

1728


Fig. 3.

Deployment description.

• TL: TL is a magnitude that indicates the effects of a variety of propagation phenomena in the underwater acoustic channel. The TL caused by spreading and absorption (or attenuation) can be calculated using the following [20]: −3

T L = 10 log r + αr × 10

f2 f2 +40 +2.75×10−4 f 2 +0.003 2 1+f 4100+f 2

⎧ (x − x)2 + (y1 − y)2 + (z1 − z)2 ⎪ ⎪ ⎨ 1 (x2 − x)2 + (y2 − y)2 + (z2 − z)2 ⎪ (x3 − x)2 + (y3 − y)2 + (z3 − z)2 ⎪ ⎩ (x4 − x)2 + (y4 − y)2 + (z4 − z)2

(3)

where α is the absorption coefficient in decibels per kilometer, and r is the range expressed in yards. At the low frequencies (100–3 kHz), the absorption coefficient can be calculated using the following [20]: α(f) = 0.1

(RSSI) technique, which uses the relationship between the transmitted power and the received power of acoustic signals to express the distance between two communication nodes [21], is expected to be more suitable for UASNs than time-of-arrival, time-difference-of-arrival, and angle-of-arrival (AOA) techniques [22]. The RSSI technique features less communication overhead and lower implementation complexity, which are particularly suitable for UASNs that typically have a severe power limit constraint. Note that, due to disparity between the TL and its measured version [see (5)], localization errors are produced when ordinary nodes try to estimate distances to anchor nodes during the localization process. The multilateration method is used to calculate coordinates of ordinary nodes. Letting (xi , yi , zi ), i = 1, 2, 3, 4, (x, y, z), denote the coordinates of four anchor nodes, some ordinary node, respectively; we then have

(4)

(6)

where di , i = 1, 2, 3, 4, denotes the distances between the ordinary node and four anchor nodes, respectively. Subtracting the fourth equation from the first three equations, we can obtain the following: ⎧ 2(x1 − x4 )x + 2(y1 − y4 )y + 2(z1 − z4 )z ⎪ ⎪ ⎪ ⎪ = x21 − x24 + y12 − y42 + z12 − z42 + d24 − d21 ⎪ ⎨ 2(x2 − x4 )x + 2(y2 − y4 )y + 2(z2 − z4 )z ⎪ = x22 − x24 + y22 − y42 + z22 − z42 + d24 − d22 ⎪ ⎪ ⎪ ⎪ 3 − y4 )y + 2(z3 − z4 )z ⎩ 2(x3 −2 x4 )x2 + 2(y = x3 − x4 + y32 − y42 + z32 − z42 + d24 − d23

where α is the absorption coefficient measured in decibels per kilometer, and f is the frequency in kilohertz. The measured TL, however, will be affected by complicating factors, such as multiple path propagation and refraction effects, which can be lumped into a parameter A, named as TL anomaly measured in decibels. To be more precise, the measured TL can be written as follows: T L = 10 log r + αr × 10−3 + A.

= d21 = d22 = d23 = d24

(7)

which can be further simplified as follows: (5)

As opposed to other polyhedron deployments, the cube and regular tetrahedron deployments will be examined in this paper since the cube and the regular tetrahedron are simplest polyhedrons and much easier to deploy in an underwater environment. This can be further illustrated using Fig. 3. Suppose √ that coordinates of A, B, and C in Fig. 3 are (0, 0, 0), (( 3/2)a, a/2, 0), and (0, a, 0), respectively. G is the √ center of gravity of ABC. Thus, the coordinate of G is (( 3/6)a, a/2, 0). At the beginning stage of the deployment, four anchor nodes are positioned at A, B, C, and G. Since underwater nodes are equipped with floating buoys that can be inflated by pumps to adjust their depths, the anchor node at G can be √ elevated to D, √ whose coordinate can be readily computed as (( 3/6)a, a/2, 6/3a). Now, the four anchor nodes form a regular tetrahedron after a simple adjustment. Considering the propagation delay and bit error rate, localization algorithms in UASNs are expected to avoid excessive overhead and use the least possible messages. Moreover, sensor nodes are constantly moving due to many environment factors; the network topology may change unpredictably as time goes on. Thus, the so-called received signal strength indicator

a 1 x + b1 y + c 1 z = s 1 a 2 x + b2 y + c 2 z = s 2 a 3 x + b3 y + c 3 z = s 3

(8)

where a1 = 2(x1 − x4 ) a2 = 2(x2 − x4 ) a3 = 2(x3 − x4 )

b1 = 2(y1 − y4 ) b2 = 2(y2 − y4 ) b3 = 2(y3 − y4 )

c1 = 2(z1 − z4 ) c2 = 2(z2 − z4 ) c3 = 2(z3 − z4 )

s1 = x21 − x24 + y12 − y42 + z12 − z42 + d24 − d21 s2 = x22 − x24 + y22 − y42 + z22 − z42 + d24 − d22 s3 = x23 − x24 + y32 − y42 + z32 − z42 + d24 − d23 . Simple computations then yield ⎧ ⎪ ⎪ ⎨x = y= ⎪ ⎪ ⎩z =

a3 (b2 c3 −b3 c2 )s1 +a3 (b3 c1 −b1 c3 )s2 +a3 (b1 c2 −b2 c1 )s3 (a3 b1 −a1 b3 )(a3 c2 −a2 c3 )−(a3 b2 −a2 b3 )(a3 c1 −a1 c3 ) a3 (a3 c2 −a2 c3 )s1 +a3 (a1 c3 −a3 c1 )s2 +a3 (a2 c1 −a1 c2 )s3 (a3 b1 −a1 b3 )(a3 c2 −a2 c3 )−(a3 b2 −a2 b3 )(a3 c1 −a1 c3 ) a3 (a2 b3 −a3 b2 )s1 +a3 (a3 b1 −a1 b3 )s2 +a3 (a1 b2 −a2 b1 )s3 (a3 b1 −a1 b3 )(a3 c2 −a2 c3 )−(a3 b2 −a2 b3 )(a3 c1 −a1 c3 ) .

Thus, by using the multilateration method, coordinates of ordinary nodes can be calculated.


Fig. 4.

Topology of the random deployment scheme.

1729

Fig. 5.

Topology of the cube deployment scheme.

Fig. 6.

Topology of the regular tetrahedron deployment scheme.

IV. S IMULATIONS In this section, we describe our parameter settings and performance evaluation criteria, and analyze and compare the performances of the three deployment schemes. A. Parameter Settings and Performance Evaluation Criteria The deployment schemes are evaluated using MATLAB in a 600 m × 600 m × 600 m monitored space. The total number of sensor nodes varies from 100 with a step size of 50 to 400 while keeping the number of anchor nodes the same. As aforementioned, anchor nodes are deployed at the vertices of each polyhedron unit, and ordinary nodes are deployed randomly in the 3-D monitored space. Both anchor nodes and ordinary nodes can adjust their pumps to float at different depths. The scenarios when the anchor node percentage are, respectively, 6.75%, 16%, and 20% are considered in our simulations. The network topologies of three deployment schemes are depicted in Figs. 4–6, respectively, where red dots represent anchor nodes, and blue circles represent ordinary nodes. For the cube deployment scheme and the regular tetrahedron deployment scheme, the side lengths of cubes and regular tetrahedrons are set to be equal to R. To ensure the reliability, the simulation results are taken as the average of 100 runs. The performances of the three deployment schemes are evaluated according to the following four criteria: localization ratio, localization error, average number of neighboring anchor nodes, and network connectivity, defined as follows. • Localization ratio is the ratio of the number of localized ordinary nodes to the total number of ordinary nodes. Obviously, the higher the localization ratio is, the more ordinary nodes can be localized. The localization ratio can be computed as follows [23]: L_ratio =

Nl No

where Nl is the number of localized ordinary nodes, and No is the total number of ordinary nodes. • Localization error is the average distance between the estimated coordinates and the real coordinates. Clearly, the smaller the localization error is, the better the localization result will be. The localization error can be computed as follows [24]: N l (ui − xi )2 + (vi − yi )2 + (wi − zi )2 L_error = i=1 Nl where (ui , vi , wi ) are real coordinates of an ordinary node i, (xi , yi , zi ) are estimated coordinates of an ordinary node i, andNl is the number of localized ordinary nodes. • Average number of neighboring anchor nodes is the ratio of the number of sensor nodes that can communicate with anchor nodes to the total number of sensor nodes. The larger the average number of neighboring anchor

1730


Fig. 7.

Localization ratio. (a) Anchor node percentage = 6.75%. (b) Anchor node percentage = 16%. (c) Anchor node percentage = 20%.

nodes is, the more choices the ordinary node will have to select appropriate anchor nodes to help with localization. The average number of neighboring anchor nodes can be computed as follows [25]: A_anchor =

Ncom_a N

where Ncom_a is the number of sensor nodes that can communicate with anchor nodes, andN is the total number of sensor nodes. • Network connectivity is the ratio of the number of sensor nodes that can communicate with other sensor nodes to the total number of sensor nodes, which can be computed as follows [26]: N _connectivity =

Ncom N

where Ncom is the number of sensor nodes that can communicate with other sensor nodes, and N is the total number of sensor nodes. B. Simulation Results 1) Localization Ratio: Fig. 7 depicts the relationship between the localization ratio and the number of sensor nodes. The regular tetrahedron deployment scheme outperforms both the random deployment scheme and the cube deployment scheme in terms of the localization ratio. When the anchor node percentage is relatively small (6.75%), the localization ratio of the random deployment scheme fluctuates more than that of the cube deployment scheme and the regular tetrahedron deployment scheme, as shown in Fig. 7(a). This is due to the fact that, in the random deployment scheme, some of the edge nodes and isolated nodes will only receive a limited number of coordinate messages from anchor nodes to help with localization. On the other hand, as the anchor node percentage increases, the fluctuation of the random deployment scheme decreases rapidly, and the localization ratios of the cube deployment scheme and the regular tetrahedron deployment scheme are nearly the same. For example, when the anchor node percentage is 16%, and the number of sensor nodes is 350, the localization ratio of the cube deployment scheme is 98.75%, and the local-

ization ratio of the regular tetrahedron deployment scheme is 100%, as depicted in Fig. 7(b), because most of the ordinary nodes can be localized when there are enough anchor nodes in the network. By comparing Fig. 7(a)–(c), we conclude that the more anchor nodes there are, the higher the localization ratio will be. For example, if the anchor node percentage is 6.75%, the localization ratio of the regular tetrahedron deployment is 86.06% when the number of sensor nodes is 150; however, if the anchor node percentage is 16%, the localization ratio of the regular tetrahedron deployment can reach 100% with the same number of sensor nodes. However, when the anchor node percentage is large enough, the localization ratio reaches a relatively high value and will not change much with the increase in the anchor node percentage. This suggests that, in sparse UASNs, we can increase the number of anchor nodes to achieve higher localization ratio. 2) Localization Error: Reducing the localization error of UASNs is one of the main motivations of this paper. Fig. 8 plots the relationship between the localization error and the number of sensor nodes. The regular tetrahedron deployment scheme has smaller localization error than the random deployment scheme and the cube deployment scheme in general. We can also observe that when the number of sensor nodes is relatively small, and the localization errors of the three deployment schemes are almost the same, particularly when the anchor node percentage is large. For example, the localization errors of the three deployment schemes are all approximately equal to 1.7 m when the number of sensor nodes is 100 and the anchor node percentage is 20%, as shown in Fig. 8(c). Fig. 8(a)–(c) suggest that the more anchor nodes there are, the smaller the localization error will be. However, the improvement is effective only when the anchor percentage is small since we only use the simplest multilateration method to calculate coordinates of ordinary nodes. Fig. 8 also reveals that the localization errors of the three deployment schemes increase with the increase of the number of sensor nodes. This is because the RSSI method uses the relationship between transmitted power and received power of wireless signals to express the distance between two communicating nodes; the longer the distance between two communication nodes, the larger the localization error caused by the RSSI-based localization technique. However, with the increase in the anchor node percentage,


Fig. 8.

1731

Localization error. (a) Anchor node percentage = 6.75%. (b) Anchor node percentage = 16%. (c) Anchor node percentage = 20%.

Fig. 9. Average number of neighboring anchor nodes. (a) Anchor node percentage = 6.75%. (b) Anchor node percentage = 16%. (c) Anchor node percentage = 20%.

the localization errors of the cube deployment scheme and the regular tetrahedron deployment scheme essentially stay the same. For instance, when anchor node percentage is 20%, the localization error of the cube deployment scheme stays around 1.78 m, and the localization error of the regular tetrahedron deployment scheme stays around 1.74 m when the number of sensor nodes varies from 100 to 400. Not surprisingly, there is a tradeoff between the localization error and the number of anchor nodes to be deployed. 3) Average Number of Neighboring Anchor Nodes: Fig. 9 shows the average number of neighboring anchor nodes of the three deployment schemes with different numbers of sensor nodes. In general, the average number of neighboring anchor nodes of the cube deployment scheme and the regular tetrahedron deployment scheme remain essentially the same in spite of the increase in the number of sensor nodes. However, the performance curve of the random deployment scheme fluctuates greatly. Therefore, the performances of the regular tetrahedron deployment and cube deployment schemes are much more stable than that of the random deployment scheme. When the anchor node percentage varies from 6.75% to 20%, the average number of neighboring anchor nodes of the cube deployment scheme barely changes. It has been observed that the regular tetrahedron deployment scheme and the random deployment scheme have more neighboring anchor nodes than the cube deployment scheme, which can be readily justified as follows: In the cube deployment scheme, anchor nodes are uniformly deployed at the vertices of each cube unit; therefore,

even if all the sides of a cube unit and a regular tetrahedron unit are of equal length, e.g., a, the maximum distance between √ two anchor nodes in a cube unit is 3a. As a result, the ordinary nodes that are closer to one anchor node may be unable to communicate with the anchor node on the diagonal of the cube, which results in a low average number of anchor nodes. The regular tetrahedron deployment scheme has more neighboring anchor nodes than the random deployment scheme when the anchor node percentage is no smaller than 16%. Therefore, ordinary nodes in the regular tetrahedron deployment scheme are more likely to help for choosing appropriate anchor nodes, which contributes to a lower localization error. 4) Network Connectivity: A lack of network connectivity may lead to an undesirable network fragmentation: isolated subnetworks may not be able to pass the information to the sinks, particularly in sparsely deployed UASNs. Fig. 10 examines the performances of the three deployment schemes in terms of network connectivity with different numbers of sensor nodes and anchor node percentages. Note that there is an obvious tradeoff between network connectivity and energy consumption: signals with high power have longer propagation distances, which entail higher network connectivity but inevitably they consume more energy. In general, network connectivity of three deployment schemes increases as the number of sensor nodes increases. When the anchor node percentage is relatively small, a larger communication range should be used to ensure the reliable

1732

Fig. 10.


Network connectivity. (a) Anchor node percentage = 6.75%. (b) Anchor node percentage = 16%. (c) Anchor node percentage = 20%.

communications between two neighboring anchor nodes. This way, more sensor nodes can communicate with others, which leads to a high network connectivity. As shown in Fig. 10, the network connectivity of three deployment schemes is similar, and the performance of the regular tetrahedron deployment scheme is slightly better than that of the other two schemes. Fig. 7–10 suggest that an anchor node percentage of 16% is ideal for improving localization ratio, reducing localization error and increasing average number of anchor nodes while maintaining a reasonable network connectivity. V. C ONCLUSION AND F UTURE R ESEARCH I SSUES In this paper, we investigated the fundamental issue of node deployment in UASNs. We proposed three deployment schemes in 3-D UASNs (the random deployment, cube deployment, and regular tetrahedron deployment schemes) and compare their performances in detail in terms of localization ratio, localization error, average number of neighboring anchor nodes, and network connectivity. Confirming that the common belief that appropriately chosen node deployment strategy is of central importance for accurate localizations in UASNs, the simulations conducted in this paper revealed that the regular tetrahedron deployment scheme outperforms the other two schemes in terms of reducing localization error and increasing localization ratio, while maintaining the average number of neighboring anchor nodes and a reasonable network connectivity. We list some possible further directions in the following. • A natural follow-up direction is an in-depth quantitative treatment of the three proposed deployment schemes since such a treatment will offer more insights regarding the design of 3-D UASNs. • A more realistic model should take into account the mobility of sensor nodes under real water conditions. • Most of the current existing node deployment algorithms assume a safe and credible environment, which, in many real applications, is barely the case. • Since underwater sensor nodes are expensive and the costs quickly rise for deep water, deployment algorithms for mobile anchor nodes are expected to play a crucial role in reducing costs in deep water scenarios.

R EFERENCES [1] H. Tan, R. Diamant, W. K. G. Seah, and M. Waldmeyer, “A survey of techniques and challenges in underwater localization,” Ocean Eng., vol. 38, no. 14/15, pp. 1663–1676, Oct. 2011. [2] R. B. Manjula and S. S. Manvi, “Issues in underwater acoustic sensor networks,” Int. J. Comput. Electr. Eng., vol. 3, no. 1, pp. 1793–8163, Feb. 2011. [3] G. Han, C. Zhang, L. Shu, N. Sun, and Q. Li, “A survey on deployment algorithms in underwater acoustic sensor networks, international journal of distributed sensor networks,” Int. J. Distrib. Sens. Netw., vol. 11, p. 2013, 2013, Art. ID 314049. [4] D. P. Williams, “AUV-enabled adaptive underwater surveying for optimal data collection,” Intell. Serv. Robit., vol. 5, no. 1, pp. 33–55, Jan. 2012. [5] N. S. N. Ismail, L. A. Hussein, and S. H. S. Ariffin, “Analyzing the performance of acoustic channel in underwater wireless sensor network (UWSN),” in Proc. Int. Conf. AMS, May 2010, pp. 550–555. [6] R. C. Luo and O. Chen, “Mobile sensor node deployment and asynchronous power management for wireless sensor networks,” IEEE Trans. Ind. Electron., vol. 59, no. 5, pp. 2377–2385, May 2012. [7] B. Huang, C. Yu, and B. D. O. Anderson, “Understanding error propagation in multihop sensor network localization,” IEEE Trans. Ind. Electron., vol. 60, no. 12, pp. 5811–5819, Dec. 2013. [8] D. Li, Z. Li, W. Ma, H. Chen, and W. Chen, “Constrained surface-level gateway placement for underwater acoustic wireless sensor networks,” Theor. Comput. Sci., vol. 447, pp. 100–106, Aug. 2012. [9] W. Alsalih, S. Akl, and H. Hassanein, “Placement of multiple mobile data collectors in underwater acoustic sensor networks,” Wireless Commun. Mobile Comput., vol. 8, no. 8, pp. 1011–1022, Oct. 2008. [10] J. Nie, D. Li, and Y. Han, “Optimization of multiple gateway deployment for underwater acoustic sensor networks,” Comput. Sci. Inf. Syst., vol. 8, no. 4, pp. 1073–1095, Oct. 2011. [11] N. Aitsaadi, N. Achirt, K. Boussettat, and G. Pujolle, “Differentiated underwater sensor network deployment,” in Proc. OCEANS, Oct. 2007, pp. 1–6. [12] C. Chang and H. Chang, “Energy-aware node placement, topology control and MAC scheduling for wireless sensor networks,” Comput. Netw., vol. 52, no. 11, pp. 2189–2204, Aug. 2008. [13] J. Kong, J. Cui, D. Wu, and M. Gerla, “Building underwater ad-hoc networks and sensor networks for large scale real-time aquatic application,” in Proc. IEEE MILCOM, Oct. 2005, pp. 1535–1541. [14] J. Cui, J. Kong, M. Gerla, and S. Zhou, “Challenges: Building scalable mobile underwater wireless sensor networks for aquatic applications,” IEEE Netw., vol. 20, no. 3, pp. 12–18, May/Jun. 2006. [15] P. Xie, J. Cui, and L. Lao, “VBF: Vector-based forwarding protocol for underwater sensor networks,” in Proc. Int. Conf. NETWORKING, May 2006, pp. 1216–1221. [16] W. K. G. Seah and H. X. Tan, “Multipath virtual sink architecture for underwater sensor networks,” in Proc. Asia Pacific OCEANS, May 2006, pp. 1–6. [17] S. M. N. Alam and Z. J. Haas, “Coverage and connectivity in three-dimensional networks,” in Proc. IEEE MOBICOM, Sep. 2006, pp. 346–357. [18] S. M. N. Alam and Z. J. Haas, “Coverage and connectivity in threedimensional underwater sensor networks,” Wireless Commun. Mobile Comput., vol. 8, no. 8, pp. 995–1009, Oct. 2008. [19] K. Akkaya and A. Newell, “Self-deployment of sensors for maximized coverage in underwater acoustic sensor networks,” Comput. Commun., vol. 32, no. 7–10, pp. 1233–1244, May 2009.


[20] M. C. Domingo and R. Prior, “Energy analysis of routing protocol for underwater wireless sensor networks,” Comput. Commun., vol. 31, no. 6, pp. 1227–1238, Apr. 2007. [21] G. Han, H. Xu, T. Q. Duong, J. Jiang, and T. Hara, “Localization algorithms of wireless sensor networks: A survey,” Telecommun. Syst., vol. 52, no. 4, pp. 2419–2436, Apr. 2013. [22] Y. Liu, Z. Yang, X. Wang, and L. Jian, “Location, localization, localizability,” J. Comput. Sci. Technol., vol. 25, no. 2, pp. 274–297, Mar. 2010. [23] H. Cui and Y. Wang, “Four-mobile-beacon assisted localization in threedimensional wireless sensor networks,” Comput. Electr. Eng., vol. 38, no. 3, pp. 652–661, May 2012. [24] B. S. Choi, J. W. Lee, J. J. Lee, and K. T. Park, “A hierarchical algorithm for indoor mobile robot localization using RFID sensor fusion,” IEEE Trans. Ind. Electron., vol. 58, no. 6, pp. 2226–2235, Jun. 2011. [25] H. Cho, J. Lee, D. Kim, and S. W. Kim, “Observability-based selection criterion for anchor nodes in multiple-cell localization,” IEEE Trans. Ind. Electron., vol. 60, no. 10, pp. 4554–4561, Oct. 2013. [26] F. M. A. Turjman, H. S. Hassanein, and M. Ibnkahla, “Quantifying connectivity in wireless sensor networks with grid-based deployments,” J. Netw. Comput. Appl., vol. 36, no. 1, pp. 368–377, Jan. 2013.

Guangjie Han (S’03–M’05) received the Ph.D. degree from Northeastern University, Shenyang, China, in 2004. From 2004 to 2006, he was a Product Manager for the ZTE Company. In February 2008, he finished his work as a Postdoctoral Researcher with the Department of Computer Science, Chonnam National University, Gwangju, Korea. From October 2010 to 2011, he was a Visiting Research Scholar with Osaka University, Suita, Japan. He is currently a Professor with the Department of Information and Communication System, Hohai University, Nanjing, China. He is the author of over 130 papers published in related international conference proceedings and journals, and is the holder of 55 patents. His current research interests include sensor networks, computer communications, mobile cloud computing, and multimedia communication and security. Dr. Han has served as a Cochair for more than 20 international conferences/workshops and as a Technical Program Committee member of more than 70 conferences. He has served on the Editorial Boards of up to 14 international journals, including the Journal of Internet Technology and KSII Transactions on Internet and Information Systems. He has served as a Reviewer of more than 50 journals. He received the 2014 Second International Conference on Computing, Management and Telecommunications and International Conference on Communications and Networking in China Best Paper Awards. He is a member of the Association for Computing Machinery.

Chenyu Zhang received the B.S. degree in communication engineering from Hohai University, Nanjing, China, in 2012. She is currently working toward the Master’s degree in the Department of Communication and Information Systems, Hohai University, Nanjing, China. Her current research interests include node deployment and path planning for sensor networks.

Lei Shu (M’07) received the Ph.D. degree from the National University of Ireland, Galway, Ireland, in 2010. Until March 2012, he was a Specially Assigned Researcher with the Department of Multimedia Engineering, Graduate School of Information Science and Technology, Osaka University, Japan. Since October 2012, he has been with Guangdong University of Petrochemical Technology, Maoming, China as a Full Professor. Since 2013, he has been a Ph.D. Supervisor with Dalian University of Technology, Dalian, China, and

1733

a Master Supervisor with Beijing University of Posts and Telecommunications, Beijing, China. He has also been the Vice Director of the Guangdong Provincial Key Laboratory of Petrochemical Equipment Fault Diagnosis, Guangdong University of Petrochemical Technology. He is the Founder of the Industrial Security and Wireless Sensor Networks Laboratory. He is the author of over 200 papers published in related conference proceedings, journals, and books. His current H-index is 18. His research interests include wireless sensor networks, multimedia communication, middleware, security, and fault diagnosis. Dr. Shu served as a Cochair for more than 50 various international conferences/workshops, e.g., the IEEE International Wireless Communications and Mobile Computing Conference (IWCMC), the IEEE International Conference on Communications (ICC), the IEEE Symposium on Computers and Communications (ISCC), the IEEE International Conference on Computing, Networking and Communication, and the International Conference on Communications and Networking in China (Chinacom). He also served/will serve as a Symposium Cochair for IWCMC 2012 and ICC 2012, as a General Chair for Chinacom 2014 and the 2015 International Conference on Heterogeneous Networking for Quality, Reliability, Security, and Robustness (Qshine), as a Steering Chair for the 2015 International Conference on Industrial Networks and Intelligent Systems, and as Technical Program Committee members of more than 150 conferences, including the IEEE International Conference on Distributed Computing in Sensor Systems, the IEEE International Conference on Mobile Ad hoc and Sensor Systems, ICC, Globecom, IEEE International Conference on Computer Communications and Networks, IEEE Wireless Communications and Networking Conference, and ISCC. He currently serves as the Editor-in-Chief for the European Alliance for Innovation Endorsed Transactions on Industrial Networks and Intelligent Systems, and the Associate Editor for a number of renowned international journals. He received the 2010 IEEE Global Communications Conference and 2013 IEEE International Conference on Communications Best Paper Awards. He is a member of the IEEE Communication Society, the European Alliance for Innovation, and the Association for Computing Machinery.

Joel J. P. C. Rodrigues (S’01–M’06–SM’06) received the B.Sc. degree (licentiate) in informatics engineering from the University of Coimbra, Coimbra, Portugal, and the M.Sc. degree in informatics engineering and the Ph.D. degree from the University of Beira Interior, Covilh, Portugal. He is currently a Professor with the Department of Informatics, University of Beira Interior, and a Researcher with the Instituto de Telecomunicaes, Lisbon, Portugal. He is also with Saint Petersburg State University of Information Technologies, Mechanics and Optics, Saint Petersburg, Russia. He is the author or coauthor of over 350 papers published in refereed international journals and conference proceedings, a book, and three patents. His main research interests include sensor networks, e-health, e-learning, vehicular delaytolerant networks, and mobile and ubiquitous computing. Dr. Rodrigues is the Leader of the NetGNA Research Group (http:// netgna.it.ubi.pt), the Chair of the IEEE Communications Society (ComSoc) Technical Committee on e-health, the Past Chair of the IEEE ComSoc Technical Committee on Communications Software, a Member Representative of the IEEE ComSoc on the IEEE Biometrics Council, and an Officer of the IEEE 1907.1 Standard. He is a member of many international technical program committees and has participated in organizing several international conferences. He is the Editor-in-Chief of the International Journal on E-Health and Medical Communications, the Editor-in-Chief of Recent Advances on Communications and Networking Technology, and an Editorial Board member of several journals. He has been a General Chair and a Technical Program Committee Chair of many international conferences. He received the Outstanding Leadership Award at the 2010 IEEE Global Communications Conference as a Communication Software, Services and Multimedia Applications Symposium Cochair and several best papers awards. He is a Licensed Professional Engineer (as a senior member), a member of the Internet Society, a Fellow of the International Academy Research and Industry Association, and a Senior Member of the Association for Computing Machinery.