Node Self-Deployment Algorithm Based on an Uneven Cluster with Radius Adjusting for Underwater Sensor Networks

Transcription

1 sensors Article ode Self-Deployment Algorithm Based on an Uneven Cluster with Radius Adjustg for Underwater Sensor etworks Peng Jiang 1,2, *, img Xu 1,2 Feng Wu 1,2 Received: 21 ovember 2015; Accepted: 7 January 2016; Publhed: 14 January 2016 Academic Editor: Jaime Lloret Mauri 1 College Automation, Hangzhou Dianzi University, Hangzhou , Cha; xymhdu@163.com (.X.); fengwu@hdu.edu.cn (F.W.) 2 Key Lab for IOT Information Fusion Technology Zhejiang, Hangzhou , Cha * Correspondence: pjiang@hdu.edu.cn; Tel.: (ext. 512); Fax: Abstract: Extg move-restricted self-deployment algorithms are based on a fixed communication radius, evaluate performance based on network coverage or connectivity rate do not consider number s near sk energy consumption dtribution network topology, reby degradg network reliability energy consumption balance. Therefore, we propose a dtributed underwater self-deployment algorithm. First, each begs uneven g based on dtance on water surface. Each selects its next-hop to synchronously construct a connected path to sk. Second, adjusts its depth while matag layout formed by uneven g n adjusts positions - s. The algorithm origally considers network reliability energy consumption balance durg deployment considers coverage redundancy rate all positions that a may reach durg position adjustment. Simulation results show, compared to connected domatg set (CDS) based depth computation algorithm, that proposed algorithm can crease number s near sk improve network reliability while guaranteeg network connectivity rate. Moreover, it can balance energy consumption durg network operation, furr improve network coverage rate reduce energy consumption. Keywords: self-deployment; uneven g; radius adjustg; network reliability 1. Introduction Related Works Underwater wireless sensor networks (UWSs) are network-monitorg systems constg sensor s with capabilities perception, acoustic communication computg an underwater environment a self-organized manner; ir function to transmit sensed formation to a sk for processg analys [1,2]. UWSs can be applied to water environment monitorg, underwater exploration, mare military defense or fields [3 5]. Th type network has become one most significant topics formation field. Research on UWSs maly volves deployment, localization, time synchronization, network protocol design, etc. [6 9]. Among se topics, deployment, as foundation work for UWSs design, has a direct effect on quality target coverage network service. The -deployment problem refers to movg sensor s to ir positions an artificial or a self-organized manner to form a network topology that has special charactertics can benefit future works [10]. Accordg to mobility capability s, deployment can be divided to static deployment, free-to-move self-deployment move-restricted Sensors 2016, 16, 98; doi: /s

2 Sensors 2016, 16, self-deployment [11 16]. Static deployment assumes that s do not have capability to move need to use artificial methods to deploy [17]. It usually a centralized algorithm, needg a priori environment formation to calculate pre-deployment positions; thus, it appropriate many UWS cases, such as underwater resource exploration mare military defense [4]. Free-to-move self-deployment assumes that s have capability to move freely can move all directions [2]. Move-restricted self-deployment assumes that s have capability to move only vertically can adjust ir depth by mselves [18]. These two algorithms do not need any a priori formation, so y are applied more widely. However, former, needg some autonomous underwater vehicles to asst durg deployment process, leads to high energy resource costs, as well as high operation difficulty [16,19,20]. On contrary, latter, havg developed a simple underwater sensor with vertical mobility by method fillg or drawg f water, operates simply availably [21,22]. Thus, move-restricted self-deployment more practical than free-to-move deployment. On one h, a number related studies have been conducted on method move-restricted self-deployment. Wu et al. [23] proposed a Voronoi-based depth-adjustment scheme for UWSs. To maximize network coverage rate, th scheme compares Voronoi region area every same layer to determe s that should go down to next layer. Th process contues layer by layer until s last layer have been determed. However, th method uses a centralized manner to adjust depth, which difficult to achieve practice. Akkaya et al. [24] proposed a dtributed self-deployment depth-adjustment algorithm to adjust depth s after ir itial deployment to reduce coverage overlaps between two neighborg s. odes contue to adjust ir depth until sensor coverage can no longer be improved. Du et al. [25] proposed a coverage algorithm based on fixed-directional movement for underwater sensor networks; th algorithm uses virtual forces to adjust a to a position with zero jot force exertion achieves telligent deployment an dependent manner. However, aforementioned algorithms all view network coverage rate as a stard ignore effect network connectivity rate on network quality service. Senel et al. [26] proposed connected domatg set (CDS) based depth computation algorithm (CDA) on bas [24]. The algorithm itially constructs a connected backbone n uses domator on backbone to dividually optimize position its non-domators iteratively. The algorithm ensures full-network connectivity while maximizg network coverage reducg network deployment consumption. However, similar to or algorithms, CDA considers fixed communication radius s as deployment objects, which makes s not be able to select an appropriate position flexibly durg process depth adjustment. CDA also views network coverage rate or network connectivity rate as criterion to deploy s does not consider number s deployed with a certa range around sk. Those two crease network energy consumption decrease network performance. In addition, algorithm (i.e., CDA) considerg both network coverage rate network connectivity rate, only considers redundant coverage rate location where dtance between its next-hop R c durg process optimizg position; thus, network coverage rate can still be improved. Furrmore, energy consumption balance network formed by CDA performs poorly when number s relatively sparse. On or h, some studies [27,28] have also vestigated network formation wireless sensor networks (WSs). Hezelman et al. [29] proposed low-energy adaptive g hierarchy (LEACH) algorithm. Th algorithm selects a romly completes g with prciple nearest neighbor. Its operation simple, but it cannot control dtribution effectively. Mao et al. [30] proposed an energy-efficient g scheme (EECS), which selects a accordg to residual energy n fhes network. Compared to LEACH, EECS selects s with more residual energy preferentially as s, which reduces exchange frequency enhances network stability. Tsai [31] proposed a

3 Sensors 2016, 16, coverage-preservg routg protocol for WSs. It selects based on coverage redundancy rate n fhes network. Consequently, network operates based on, great network coverage rate mataed. In method, a havg a greater coverage redundancy rate has priority to become a, which reduces failure impact on network coverage rate slows down drop speed network coverage rate. Abbasi et al. [32] surveyed different formation algorithms, analyzg ir objects, features, etc. Lloret et al. [33] proposed an algorithm that can structure topology different WSs to coext same environment based on network g, where s manage ir own networks have connections with or s, resultg good network connectivity scalability whole parallel network structure. uan et al. [34] proposed an uneven mechanm. The mechanm divides network to many s with different sizes. Among m, a, furr from sk, has a greater size. On contrary, a, closer to sk has a smaller size. The dtribution may balance network energy consumption extend network lifetime. Qiao et al. [35] proposed a cha structure-based uneven routg algorithm. It establhes a dynamic multiple-hop route considerg uneven mechanm for transmittg data, which can balance energy consumption s prolong network lifetime. To solve aforementioned problems on move-restricted self-deployment accordg to aforementioned formulation algorithm description, th study proposes uneven radius-adjustg self-deployment algorithm (URSA). After s are scattered on water surface romly uniformly, each begs uneven g process accordg to dtance to sk. The s n use hybrid radius path-selection method simultaneously to form a connected path to sk. In depth-adjustment phase, s adjust ir own depths accordg to prciple matag layout formed on water surface by uneven g process. For each its - s, calculates some correspondg coverage redundancy rates (CRR) by assumg that - deployed on each position with maximum communication radius its basic s (i.e., s that belong to same that next-hop - ). The subsequently mimizes hop number - to optimize adjust its position. Th algorithm forms a network layout uneven dtribution. In dtribution, density number s area close to sk crease; scale that area becomes small. As a result, number s near sk n creases, network reliability improves. The process optimizg adjustg position - s furr improves network coverage rate, decreases hop - s decreases energy consumption network deployment. The connected topological structure formed durg process deployment (i.e., farr from sk has a greater communication radius, farr from sk has a larger scale) also balances energy consumption network operation. The simulation results show that compared to CDA havg a good deployment performance, URSA can improve network reliability, balances reduces network energy consumption improves network coverage rate. The rest th paper organized as follows. Section 2 describes system model, assumptions defitions considered th study. Section 3 presents details URSA. Section 4 analyzes complexity URSA. Section 5 dcusses performance study provides a detailed analys its result. Fally, Section 6 concludes paper plans some future works. 2. Models Defitions 2.1. etwork Model Assume that sensor s are scattered on water surface romly uniformly are floatg on water surface with buoys. Each has communication, perception mobility

4 Sensors 2016, 16, capabilities (perpendicular to horizontal direction), positions sensor s are adjusted underwater by movg m to form a 3D coverage network. A typical UWS architecture Sensors 2016, 16, 98 Sensors 2016, 16, 98 depicted 1. In th model, sk communicates with ground-monitorg station ground-monitorg by radio; s communicate station by with radio; one s anor communicate by acoustic with channels one anor mata by acoustic connectivity channels with ground-monitorg station by radio; s communicate with one anor by acoustic channels mata sk connectivity via one- with or multi-hop sk paths. Furrmore, via one- multi-hop se s paths. anchor Furrmore, to fix ir position se s once mata connectivity with sk via one- or multi-hop paths. Furrmore, se s anchor y have to fix adjusted ir position ir depth. once Th y study have denotes adjusted ir i-thdepth. by Th s i, study denotes correspondg i-th by set anchor to fix ir position once y have adjusted ir depth. Th study denotes i-th by si, S = {s 1, s 2, correspondg... s n }. The followg set assumptions {s1, s2, are sn}. considered: The followg assumptions are considered: si, correspondg set S = {s1, s2, sn}. The followg assumptions are considered: Underwater Underwater wireless wireless sensor sensor network network (UWS) (UWS) system system model. model. 1. Underwater wireless sensor network (UWS) system model. 1. The Boolean perception model adopted to describe sensg. If sensg radius 1. The Boolean perception model adopted to describe sensg. If sensg radius si Rs, space sensed by sphere whose center location si s i Rs, R s, space sensed by a sphere whose center location Rs as radius. An example 2. The sphere with radius sensed space Rs R s as radius. An example 2. The sphere with radius Rs s sensed space si. P1 with sphere can be covered by si. On contrary, beyond sphere si. s i. P1 P 1 with sphere can be covered by si. s i. On contrary, P2 2 beyond sphere cannot be covered by si. cannot be covered by si. s i. s i i P 1 P 1 R S R S P 2 P D Boolean perception model. 2. 3D Boolean perception model. 2. All s are omorphic before deployment, but can adjust ir transmsion power by 2. All s are omorphic before deployment, but can adjust ir transmsion power by 2. mselves, All s are i.e., omorphic communication before deployment, radius Rc but can adjust can be ir adjusted transmsion with an power adjustg by mselves, i.e., communication radius Rc can be adjusted with an adjustg precion mselves, i.e., m, but communication not more than radius maximum R c communication can be adjusted radius Rc_Max with an determed adjustg precion m, but not more than maximum communication radius Rc_Max determed by precion physical 1 m, device. but not The more communication than maximum radius communication will be radius changed R c by _Max determed dtance to by physical device. The communication radius will be changed by dtance to its bynext-hop physical, device. i.e., The communicationradius radius a will be rounded-down changed by value dtance to its next-hop, i.e., communication radius rounded-down value dtance its next-hop between, m i.e., when communication its next-hop radius determed. a rounded-down value dtance between m when its next-hop determed. 3. The dtance network between has only mone when sk its. next-hop Its position determed. fixed at center water, 3. The network has only one sk. Its position fixed at center water, 3. transmsion The networkpower has only one energy skcan. be fite. Its position fixed at center water, transmsion power energy can be fite. 4. transmsion knows power its own position energy can be can fite. determe dtance to source accordg to 4. A knows its own position can determe dtance to source accordg to 4. strength A knows received its own signal. position can determe dtance to source accordg to strength received signal. strength received signal ode Energy Consumption Model 2.2. ode Energy Consumption Model 2.2. ode Energy Consumption Model Underwater sensor network s communicate with one anor by acoustic signals. Underwater sensor network s communicate with one anor by acoustic signals. Accordgly, Underwater th study sensoruses network an energy sconsumption communicate model with one underwater anor sensor by acoustic network signals. data Accordgly, th study uses an energy consumption model underwater sensor network data communication Accordgly, thwith study uses sound an energy wave consumption as medium model [36]. The underwater underwater sensor acoustic networksignal data communication with sound wave as medium [36]. The underwater acoustic signal attenuation model A(d) given as follows: attenuation model A(d) given as follows: se rec se rec se rec se rec d s, s d s, s se rec Ad se rec Ad s, s d s, s (1) (1) Equation (1) describes energy attenuation when data packet transmittg dtance from Equation (1) describes energy attenuation when data packet transmittg dtance from source sse to destation srec d(sse, srec), where energy diffusion factor ( source sse to destation srec d(sse, srec), where λ energy diffusion factor (

5 Sensors 2016, 16, communication with sound wave as medium [36]. The underwater acoustic signal attenuation model A(d) given as follows: A pd ps se, s rec qq d λ ps se, s rec q α dps se,s rec q (1) Equation (1) describes energy attenuation when data packet transmittg dtance from source s se to destation s rec d(s se, s rec ), where λ energy diffusion factor ( cyldrical diffusion one; actual situation 1.5; spherical diffusion two). The parameter α = 10 a( F r )/10, where absorption coefficient a(f r ) as follows: apf r q F 2 r 1 ` F 2 r ` F 2 r 4100 ` F 2 r ` 2.75 ˆ 10 7 F 2 r ` 3 ˆ 10 6 (2) where F r carrier frequency, with unit khz. The unit absorption coefficient db/m. The energy for s to send data E tx (d(s se,s rec )) expressed as follows: E tx pd ps se, s rec qq P r ˆ T p ˆ A pd ps se, s rec qq (3) where T p data transmsion time P r mimum power packets that can be received Coverage Redundancy Rate The coverage redundancy rate (CRR) s i defed as ratio sensg overlappg area itself with or s with its one-hop communication radius its sensg area [37]. γ(s i ) formulated as follows: γps i q volume s j Pneighborps i q areaps j q volume pareaps i qq Xareaps i q (4) where area(s i ) represents sensg area s i. The formulation practical use defed as follows: γps i q 1 nź 2 1 j 1 «2 3 π R 3 s d3 ps i, s j q πdps i, s j q R 2 s d2 ps i, s j q πr3 s ff (5) where n number neighbor s s i d(s i,s j ) dtance between s i its neighbor s j etwork Coverage Rate The network coverage rate, which reflects coverage degree an underwater sensor network coverg a monitorg area or targets, also a primary stard to evaluate a -deployment algorithm. Th stard defed as ratio effective area, number UWS covers targets to entire area target region or all target numbers by usg Cor expression; formula expressed as follows [26]: Cor V covered V region (6) where V covered volume monitorg area covered by active s V region volume total monitorg area.

6 Sensors 2016, 16, etwork Connectivity Rate The network connectivity rate an important criterion service quality sensor networks preme sensor network application. Th criterion refers to ratio number s that connect with sk by one or several hops to total number s by usg Cer expression; formulation expressed as follows [26]: Cer connection all (7) where connection number s connected to sk all total number s UWS etwork Reliability In UWSs, network reliability reflects effect on network service quality when a does not function because serious environmental external factors or a lack energy. From deployment, network reliability can be expressed by redundancy a network. In th study, network reliability can be expressed by number s with a certa range sk average number ir neighbor s. If two dicators are great, network reliability also great. 3. Algorithm Description Process 3.1. Problem Description Some scholars have already researched self-deployment problem for move-restricted underwater sensor s. However, se algorithms are suitable only for deployment underwater sensor s with a fixed communication radius only consider criterion network coverage or network connectivity rate without considerg network reliability problem energy consumption balance network. In extg algorithms, adjacent s are far away from one anor as much as possible to mimize overlappg area among mselves to maximize network coverage rate. Consequently, probability that s are deployed near sk decreases to some degree, number s with a certa range around sk cannot be guaranteed. Even use a topology control method with a good effect cannot improve network performance after deployment. In addition, process adjustg depth, algorithms only focus on CRR position where overlappg area one with a specific smallest ignores CRR position where overlappg area not smallest. Therefore, network coverage rate should be improved. Moreover, extg algorithms that consider network connectivity rate (i.e., CDA) do not consider balance network energy consumption durg network operations. When number s dense, network formed by CDA can modify network topology by a routg protocol to achieve energy consumption balance. By contrast, when number s relatively sparse, network formed by CDA may have no or connected topology or effectively route protocols because number s sufficient. At th time, an operated network can only use topological structure formed durg process deployment. Furrmore, energy consumption balance topological structure performs badly, i.e., energy consumption balance network formed by CDA poor when number s relatively sparse. Therefore, th study defes followg problem: After s are romly uniformly scattered on water surface target area, a self-deployment algorithm with an adjustable communication radius designed to adjust depth s, improve network reliability, balance, reduce energy consumption, maximize network coverage on preme matag network connectivity. To solve defed problem, th study proposes a dtributed self-deployment algorithm for underwater sensor s (i.e., URSA). After all s are scattered on water surface romly

7 Sensors 2016, 16, uniformly, each firstly begs process uneven g accordg to dtance to sk. Th process forms a layout where number s an area close to sk becomes large scale that area becomes small. Second, every uses hybrid radius path-selection method to select its next-hop concurrently accordg to prciple mimizg energy consumption sendg a from itself to sk. The n forms a connected path to sk. The feature connected path that if far away from sk, its path to next-hop long, thus ensurg network connectivity savg balancg energy consumption network operations. Third, each begs to adjust depth an iterative manner, correspondg selects its own adjustg position with prciple that dtance between itself its next-hop after adjustment 1 m longer than that before adjustment. Th process retas layout uneven (i.e., number s an area close to sk creases, whereas communication radius that area decreases), creases number with scope sk, improves network reliability balances network energy consumption. Fally, for each -, calculates some correspondg CRR rates by supposg that - deployed on each position with maximum communication radius its basic s selects best position for every - by priority hop basic - on bas decreasg network coverage rate. Th process decreases hop number - s, creases chances s near sk, decreases energy consumption deployment improves total coverage rate. The detailed description presented as follows Algorithm Description URSA divided to followg four steps: (1) uneven g [38]; (2) constructg a connected path by hybrid radius path-selection method; (3) calculatg depth each a ; (4) fdg a next needg to be adjusted. The detailed steps algorithm are as follows Uneven Clusterg After s are scattered on water surface romly uniformly, sk broadcasts to all s. Every s i calculates dtance to sk d(s i,sk) accordg to strength signal it receives sets its probability threshold T h (s i ) by Equation (8). Every s i n generates a rom number r(s i ) from zero to one compares th number to T h (s i ). If r(s i ) < T h (s i ), s i becomes a provional jos set provional s P (P = {p 1,,... p n1 } = {s i r(s i ) < T h (s i ), s i ɛ S}, where p i represents i-th provional ); orwe, s i exits from competition transforms to sleep mode. $ & T h1 dps i, Skq ă d hot T h ps i q T h2 d hot ď dps i, Skq ă 2d hot (8) % T h3 dps i, Skq ě 2d hot where T h1, T h2 T h3 are probability threshold parameters, T h1 > T h2 > T h3. d hot represents radius hot-spot area. Equation (8) represents that if a close to sk, its T h great. In or words, s near sk have a great probability to become a provional, whose aim to crease number fal s near sk as much as possible. Every provional p i defes its own competition radius R(p i ) by Equation (9) broadcasts its status ( ID, d(p i,sk), R(p i )) with communication radius R 0. After a provional p i receives, it calculates dtance d(p i, p j ) to

8 Sensors 2016, 16, signal-sendg p j defes set its own competition neighbor s S CH (p i ) by Equation (10), namely provional with its competition radius R(p i ). Rpp i q ˆ 1 c d max dpp i, Skq R 0 (9) d max where d max farst dtance between any pots water surface target area sk ; c proportional constant; R 0 maximum competition radius. S CH pp i q p jˇˇdppi, p j q ď max `Rpp i q, Rpp j q, dpp i, p j q ď R 0 ( (10) In stage competition, all provional s set same waitg time (Time). Durg Time, each provional p i compares its dtance to sk Sensors with 2016, 16, 98 s S CH (p i ) determes wher it will become a fal (i.e., ) f m (f m m-th element set fal, namely with s SCH(pi) determes wher it will become a fal m-th(i.e., fal ) fm ) (fm accordg m-th element to followg set fal conditions:, namely m-th Sensors fal 2016, 16, 98 ) accordg to followg conditions: (1) d(p i,sk) ď m(d(p j,sk)), p j ɛ S CH (p i ): p i becomes f m broadcasts been (1) d(pi,sk) m(d(pj,sk)), pj ϵ SCH(pi): pi becomes fm with Ms CH to notify SCH(pi) its competition determes wher neighbor it will broadcasts s become with a fal been communication radius R 0. (i.e., MCH ) to notify fm (fm its competition m-th element neighbor s set with fal communication, radius namely R0. (2) d(p i,sk) > m(d(p j,sk)), p j ɛ S CH (p i ): p i waits for s from its competition neighbor (2) m-th Sensors d(pi,sk) fal 2016, 16, 98 > m(d(pj,sk)), ) accordg pj ϵ SCH(pi): to pi waits followg for s conditions: from its competition neighbor s s takes followg correspondg operation accordg to s it receives: (1) d(pi,sk) takes m(d(pj,sk)), followg correspondg pj ϵ SCH(pi): operation pi becomes accordg fm to s it receives: with s SCH(pi) determes wher it will broadcasts become a fal been (i.e., pi p MCH i receives ) to MCH notify fm broadcasted M(fm its CH competition broadcasted m-th by element one neighbor its by competition one s set with its neighbor fal competition communication pj, gives neighbor, radius up namely R0. p j, gives up (2) m-th d(pi,sk) fal > competition m(d(pj,sk)), ) accordg pj competition broadcasts ϵ SCH(pi): to pi waits followg exit broadcasts for competition s conditions: from exit its competition MGC. It n neighbor converts s to M GC. It n (1) d(pi,sk) takes sleep converts mode. m(d(pj,sk)), followg to sleep correspondg pj mode. ϵ SCH(pi): operation pi becomes accordg fm broadcasts to s been it receives: pi MGC by one its competition neighbor pj, removes pj pi MCH p i receives to MCH notify Mbroadcasted its competition GC broadcasted by one neighbor by its competition one s its with competition neighbor communication neighbor pj, gives radius up R0. p j, removes p j (2) d(pi,sk) from > competition SCH(pi), m(d(pj,sk)), n compares pj S dtance to sk with its updatg competition CH (p i ), nbroadcasts ϵ compares SCH(pi): pi waits exit for dtance competition s trom its sk competition MGC. It with n neighbor converts its updatg s to competition takes sleep neighbor mode. followg s s correspondg contues contues precedg operation precedg accordg operations. to operations. s it receives: pi cannot receive any Time becomes fm. pi MGC by one its neighbor pj, removes pj pi preceives i cannot MCH receive broadcasted any by one its competition Time neighbor becomes f m. pj, gives up After from Time competition SCH(pi), over, n every compares broadcasts dtance exit fm broadcasts to competition sk elect with its MGC. updatg It n competition converts ( to ID After fm) Time with neighbor sleep mode. over, maximum s every communication contues precedg radius Rc_Max foperations. m broadcasts to stimulate elect non- s ( ID f to jo one s to save formation its neighbor m ) with pi cannot receive any Time becomes fm. pi receives maximum MGC broadcasted communication by one radius its competition R c _Max neighbor to stimulate pj, removes non- pj s to s. When fj receives that, it saves dtance d(fm,fj) jo one places After from ID Time s SCH(pi), fm over, to n every set compares its neighbor dtance fm broadcasts to tosk save s CH(fj); elect with formation when its updatg non- competition its neighbor ( s. ID When receives fm) with neighbor that maximum, s it communication contues calculates f j receives precedg dtance radius Rc_Max to that operations. every, to stimulate i saves who non- source dtance s that d(f m,f j ) places to IDjo one f replies s with jo (non- to save formation ID) to its nearest its neighbor m pi to cannot receive set any its neighbor Time becomes s fm. CH (f j ); when non-. s. For When non- s that do fj not receives receive that any, it y saves broadcast dtance with Rc_Max d(fm,fj) receives jo places mselves After that ID Time, to fm over, it to every calculates nearest set its neighbor dtance accordg fm to broadcasts to every reply s ir CH(fj); elect neighbor who when s. non- They source transmit ( that ID replies jo receives fm) with that maximum, jo to it communication calculates (non- by dtance radius usg Rc_Max to every neighbor to stimulate ID) as who to non- its termediate nearest source. s that. For to non- jo The one flow replies chart s with Section s jo that do not receive (non- to 3 any with save, formation si as ID) an y to example. its broadcast nearest its neighbor with R c _Max jo. s. For When non- s that do fj not receives receive that any, it y saves broadcast dtance with Rc_Max d(fm,fj) mselves to nearest accordg to reply ir neighbor s. They transmit jo places mselves ID to fm to nearest set its neighbor accordg ode to calculates reply d(si,sk),sets s ir CH(fj); neighbor when s. non- They transmit jo Th(si) produces rom jo receives that to, to it calculates by dtance by usg usg number to r(si) every neighbor from neighbor 0 to 1 as who as termediate source termediate. that. The The flow flow chart replies with Section Section jo (non- 3 with 3 with si as ID) an to example. its s i nearest as an example. R(si) < Th(si)?. For non- s that do not receive any, y broadcast with Rc_Max jo mselves to nearest accordg ode to si calculates reply d(si,sk),sets ir neighbor s. They transmit ode si becomes Th(si) pi,n produces it calculates rom R(pi) SCH(pi), sets number constant r(si) Time from 0 to actual 1 time At = 0; jo to by usg neighbor as termediate. The flow chart Section with si as an example. Broadcast MEC with communication radius R0 The gives up competition turns to sleep mode until selectg ends Broadcast MEC with communication radius R0 non-- s selects nearest itself, after process competition ends The gives up competition turns to sleep mode until selectg ends Broadcast MEC with communication radius R0 non-- s selects nearest itself, after process competition ends R(si) At = < At Th(si)? + 1 ode si calculates d(si,sk),sets ode si becomes d(pi,sk) pi,n < m(d(pj,sk)), Th(si) it calculates R(pi) SCH(pi), sets constant pjϵsch(pi) produces rom number r(si) Time from 0 to actual 1 time At = 0; R(si) At = < At Th(si)? + 1 Receive MCH from its competition neighbor pj or not ode si becomes d(pi,sk) pi,n < m(d(pj,sk)), it calculates R(pi) SCH(pi), sets constant pjϵsch(pi) Time actual time At = 0; Receive MEC from its competition neighbor pj or not At = At + 1 Receive MCH from its competition neighbor pj or not d(pi,sk) remove < m(d(pj,sk)), pj from SCH(pi) pjϵsch(pi) Receive MEC from its competition neighbor At > Time? pj or not Receive MCH from its competition neighbor pj or not pi becomes fal fm remove pj from SCH(pi) End Receive MEC from its competition neighbor At > Time? pj or not 3. Flow chart Section

9 receives that, it calculates dtance to every who source that replies with jo (non- ID) to its nearest. For non- s that do not receive any, y broadcast with Rc_Max jo mselves to nearest accordg to reply ir neighbor s. They transmit jo to by usg neighbor as termediate. The flow chart Section with si as an example. Sensors 2016, 16, ode si calculates d(si,sk),sets Th(si) produces rom number r(si) from 0 to 1 R(si) < Th(si)? ode si becomes pi,n it calculates R(pi) SCH(pi), sets constant Time actual time At = 0; At = At + 1 d(pi,sk) < m(d(pj,sk)), pjϵsch(pi) Broadcast MEC with communication radius R0 The gives up competition turns to sleep mode until selectg ends non-- s selects nearest itself, after process competition ends Receive MCH from its competition neighbor pj or not Receive MEC from its competition neighbor pj or not remove pj from SCH(pi) At > Time? pi becomes fal fm End Flow Flow chart chart Section Constructg a Connected Path by Hybrid Radius Path-Selection Method When every f m does not receive reply non- s over a period time, f m will construct a connected path to sk. f m first compares d(f m,sk) to radius hot-spot area d hot. If d(f m,sk) < d hot, f m connects with sk directly; orwe, f i will fd its next-hop from set its neighbor s CH (f m ). The detailed process expressed as follows: (1) f m selects s forward direction (i.e., direction close to sk ) from s CH (f m ) by Equation (11). These s are denoted by s forward direction f m B CH (f m ): B CH p f m q f jˇˇdp fj, Skq ă dp f m, Skq, f j P CH p f m q ( (11) (2) f m selects k s nearest to itself from s B CH (f m ) as set alternative next-hop s F CH (f m ). If number s B CH (f m ) smaller than k, F CH (f m ) = B CH (f m ). If number zero, Step (3) performed; orwe, Step (4) performed. (3) f m selects k non- s whose s are not f m forward direction with its R c _Max as set alternative next-hop s F CH (f m ). Th process guarantees that communication radius f m greater than communication radius its next-hop. (4) f m selects a s j from s F CH (f m ) by mimizg sum dtance to f m dtance to sk as its next-hop next(f m ): nextp f m q s jˇˇm `dp f m, s j q ` dps j, Skq, s j P F CH p f m q ( (12) (5) If next(f m ) a non-, becomes a fds its next-hop accordg to precedg process. Orwe, entire process ends.

10 Sensors 2016, 16, Cluster Head ode Calculatg Depth Each ode a Cluster When one f m receives calculate divg depth M CD, f m optimizes its own position position its - s. After depths all s its are calculated, each dives to its position synchronously. f m itially calculates its own divg position accordg to prciple matag adjustg layout on water surface formed by uneven g. The rules are as follows: (1) f m a common (havg both last-hop next-hop ). If communication radius f m on water surface R UA (f m ) plus one greater than adjusted communication radius R A (f j ) its next-hop f j, i.e., f m satfies feature uneven g after it adjusts its depth, f m regards its adjusted communication radius as R UA (f m ) + 1 calculates its divg depth; orwe, f m regards adjusted communication radius its next-hop R A (f j ) plus one as its adjusted communication radius calculates its divg depth. Th process calculatg divg depth f m dep(f m ) formulated as follows: depp f m q depp f j q ` # a 2RUA p f m q ` 1 R UA p f m q ` 1 ą R A p f j q b`ra p f j q ` 1 2 prua p f m qq 2 R UA p f m q ` 1 ď R UA p f j q (13) where R UA (f m ) communication radius f m before it adjusts its depth. In or words, R UA (f m ) represents dtance between itself its next-hop on water surface. R A (f m ) communication radius f m after it adjusts its depth. In or words, R A (f m ) represents dtance between itself its next-hop after f m adjusts its depth. (2) If f m a leaf (havg only next-hop s), it can reduce CRR with its last-hop by creasg its own depth. f m determes its depth dep(f m ) by Equation (14) on bas wher it one-hop sk. $ dep p f m q dep ` & f j ` % b d 2 hot d2 p f m, Skq bp2r d p f m, Skq ď d hot (14) s q 2 pr UA p f m qq 2 d p f m, Skq ą d hot After f m defes its depth, it updates set all_deployed constg s that have adjusted ir depths entire network set part_deployed constg s with depths calculated begs to calculate depth its - s. The f m defes depth one - f every round. In each round, f m itially selects each order from s part_deployed as basic b f (i.e., s same with f, as next-hop s f ). For every basic b, f m dividually calculates divg depth f by Equations (15) (16) when dtance d_adjust between adjusted position f adjusted position b R c _Max, R c _Max-1... d( f,b). f m n selects mimum CRR as CRR basic b saves correspondg depth (when mimum CRR corresponds to several different depths, f m selects correspondg depth by communication radius f divg dtance f order). b d_vertical d_adjust 2 d 2 p f, bq (15) $ & R s deppbq d_vertical ă R s depp f q Depth R s deppbq ` d_vertical ą Depth R s (16) % deppbq d_vertical ors where d_vertical vertical dtance between adjusted position f adjusted position b; Depth depth target area. Equation (16) represents divg depth f when divg depth f greater than Depth-R s or smaller than R s. In or words, at that moment, some useless region exts sensg area f, divg depth f equal to Depth-R s or R s.

11 Sensors 2016, 16, f m subsequently compares CRR every basic b to select mimum CRR. Thereafter, f m compares hop number every basic to CRR that with rang range mimum CRR selects basic with mimum hop number as basic f (when hop number equal, f m selects it by communication radius divg depth order f ). f m determes optimal position f. f m fally updates part_deployed all_deployed. An example for process by which calculates depths its - s described 4a. In figure, s 1, s 2, s 3, s 4, s 5 f are on water surface. Among m, f represents a, ors represent - s. In addition, s 1, s 2, s 3 f are part_deployed. In or words, ir depths have been already calculated. When f calculates depth s 5, it fds that dtance between s 5 each part_deployed greater than R c _Max. Thus, it ignores calculatg depth s 5 tentatively contues to calculate depth anor (i.e., s 4 ). It firstly selects s 1 as basic s 4 calculates possible divg depth assumg that dtance between adjusted position s 4 s 1 ' respectively R c _Max, R c _Max-1... d(s 4,s 1 ). Consequently, all possible divg depths are on le segment d 1 d 2. It n calculates CRR every position on d 1 d 2 chooses mimum CRR as CRR s 1 saves correspondg position m 3. In a similar way, all possible divg depths are on le segment c 1 c 2, position mimum CRR m 2, when it selects s 3 as basic s 4 ; all possible divg depths are on le segment a 1 a 2, position mimum CRR m 1, when it selects f. Specifically, when it selects s 2 as basic s 4, it ignores situation because dtance between s 4 s 2 greater than R c _Max. Fally, f fds mimum CRR among se position (i.e., m 1, m 2, m 3 ) compares hop basic whose CRR with rang range mimum CRR (i.e., compares hop s 1 f ). Consequently, f Sensors 2016, 16, 98 selects m 1 as divg depth s 4 f as basic s 4. After fhg calculatg depth Consequently, f selects m1 as divg depth s4 f as basic s4. After fhg s 4, it calculates s 5 calculatg aga. Atdepth last, s4, each it calculates synchronously dives to position accordg to s5 aga. At last, each synchronously dives to position accordg to calculation result. The result s depth adjustg depicted calculation result. The result s depth adjustg depicted 4b. 4b. s 1 f s 4 a1 s5 s 2 s3 m1 f' c1 a2 m2 s 2' s 1' d1 c2 m3 d2 s 3' Shadow itial position adjusted position (a) s2' s1' f' s4' s3' s5' (b) 4. (a) Depth calculation process a (assume that CRR on m3 mimum CRR on m1 with rang range mimum CRR; basic (next-hop ) s2 s3 f, basic s1 s2, after ir depth already calculated); (b) 3D result s depth adjustg. 4. (a) Depth calculation process a (assume that CRR on m 3 mimum CRR on m 1 with rang range mimum CRR; basic (next-hop ) s 2 s 3 f, basicthe flow chart s 1 Section s 2, after ir depth 5 with already a calculated); fm as an example. (b) 3D result s depth adjustg Fdg a ext Cluster eedg to Adjust After fm calculates divg depth all s its, it transmits adjust depth (divg depth, basic ID) to correspondg - to notify m to dive. After all s dive to specified location, fm broadcasts update (ID, divg depth adjusted communication radius every ), or s that receive update ir all_deployed. fm n queries its last-hop s wher some s have not adjusted. Thus, fm transmits MCD to one m, receivg operates by Section Orwe, fm transmits been adjusted MAD to its next-hop fj, fj selects to transmit MCD or MAD

12 Sensors 2016, 16, The flow chart Section with a f m as an example Fdg a ext Cluster eedg to Adjust After f m calculates divg depth all s its, it transmits adjust depth (divg depth, basic ID) to correspondg - to notify m to dive. After all s dive to specified location, f m broadcasts update (ID, divg depth adjusted communication radius every ), or s that receive update ir all_deployed. f m n queries its last-hop s wher Sensors 2016, 16, 98 some s have not adjusted. Thus, f m transmits M CD to one m, receivg operates by Section Orwe, f m transmits been adjusted accordg to situation its last-hop. Th process repeated until all s M AD to its next-hop f j, f j selects to transmit M CD or M AD accordg to situation its are adjusted. last-hop. Th process repeated until all s are adjusted. Start f m calculates its own depth by Equation (13) or (14) updates all_deployed part_deployed f m selects unadjusted f from s its Select b order as basic from part_deployed d( f,b) > R c _Max Calcultates f ś divg position correspondg CRR,when dtance f adjusted position b R c _Max,R c _Max-1,,,d( f,b) For current basic b, select m CRR, sign as CRR b, save correspondg depth All s part_deployed are selected or not? Among CRR every basic, calculates m CRR Selects best position for f priority to hop number its basic which with rang range m CRR All s are calculated for ir depth or not? Fd a next needg to be adjusted End 5. Flow chart Section Flow chart Section Algorithm Analys 4.1. Message Flow between odes In URSA, flow between s takes place durg process uneven g,

13 Sensors 2016, 16, Algorithm Analys 4.1. Message Flow between odes In URSA, flow between s takes place durg process uneven g, constructg a connected path position adjustment - s. Th paper illustrates flow between s algorithm based on some specific s Message Flow durg Uneven Clusterg Sensors 2016, 16, 98 Sensors There 2016, 16, 98 flow several stages uneven g, as 6. p 1 p 1 p 6 p 6 p 6 Competition p 6 Competition Competition starts Competition ends starts ends p 4 p5 p 1 p 4 p5 p 1 p 4 p5 p 1 p 4 p5 p 1 p 3 p 3 p 3 p 3 (a) (b) (c) (a) (b) (c) p 3 p 3 p 6 p 6 p 4 p 5 p 4 p 5 Sk provional fal sleep givg up competition Sk provional fal sleep givg up competition d hot d hot Several stage statuses uneven g. (a) Initialization phase 6. Several stage statuses uneven g. (a) Initialization phase competition; (b) Phase competition; (c) End competition. competition; (b) Phase competition; (c) End competition. (Assume that that, p2, p 3 p3 p 4 can p4 communicate can communicate with each with or each with or R 0 with that R0 y are that y competition are (Assume that p2, p3 p4 can communicate with each or with R0 that y are neighbor competition s neighbor for eachs or; for soeach do p 4 or;, p 5 so pdo 6. p4, In addition, p5 p6. pin 2 addition, next-hop p2 next-hop p 6, competition neighbor s for each or; so do p4, p5 p6. In addition, p2 dtance p6, between dtance p next-hop 1 between sk p1 sk greater than greater R c _Max). than Rc_Max). p6, dtance between p1 sk greater than Rc_Max). In In In 6a, 6a, 6a, s s s are are are itialization itialization itialization phase phase phase competition. competition. competition. At At At th th th time, time, time, every every every provional provional provional broadcasts broadcasts broadcasts with with with R R0; R0; 0 ; process process process flow flow flow between between between s s s 7a. 7a. 7a. Then, Then, Then, enterg enterg enterg to to to phrase phrase phrase competition competition competition 6b. 6b. p2 6b. p p2 2 firstly firstly firstly becomes becomes becomes a a fal fal fal broadcasts broadcasts MCH broadcasts M MCH CH with with with R R0, R0, 0, p p3, p3, 3, as as as well well well as as p4 as p p4 4 exit exit exit competition competition competition due due due to to to receivg receivg receivg MCH M from p2 MCH CH from from p p2 2 broadcast broadcast broadcast MGC, MGC, M respectively, GC respectively,, respectively, which which which 7b. 7b. 7b. Afterwards, p5 Afterwards, Afterwards, p p6 p5 5 p respectively p6 6 respectively respectively remove remove remove p4. p p4. 4. At At At same same same time, time, time, p6 p p6 6 becomes becomes becomes a a fal fal fal,,, p5 p 5 exits exits p5 exits competition; competition; competition; flow flow flow 7c. After 7c. After 7c. After competition, competition, competition, 6c, 6c, 6c, fal fal fal s broadcast s broadcast s broadcast ir ID, ir ir ID, ID, non- non- non- s reply s s with reply reply ir with with ID to ir ir ID correspondg ID to correspondg to correspondg,,, flow which flow flow which which 7d. 7d. 7d. M 2 M M 2 4 M 4 M4 M4 M M 5 5 M 2 M 2 3 M 3 p 4 p 4 M 4 M M 4 6 M 6 M 4 M3 M3 M 4 p 3 p 3 M f M f M f M f M M e e p 4 p 4 M p 5 p 5 p 5 M 6 p p 5 p 5 M M f :M CH (Been ) M 6 M i :p i s ID,d(p i,sk),r(p i ) M f :M e :M CH GC (Been (Exit competition) ) M i :p i s ID,d(p i,sk),r(p i ) M e :M GC (Exit competition) M e Me Me M e M e M e p 3 p 3 p 3 p 3 p 6 p 5 p 6 p 5 M M e e p 4 p 4 p 6 M f p 6 M M f f :M CH (Been ) MM f :M e :M CH GC (Been (Exit competition) ) M e :M GC (Exit competition) p 1 p 1 p 6 p 6 p 1 's ID p 1 's ID Jo(ID) Jo(ID) 's ID nu- pjo(id) 2 's nu- Jo(ID) p6's ID p6's ID Jo(ID) Jo(ID) 's ID 's ID p 6 's ID p 6 's ID (a) (b) (c) (d) (a) (b) (c) (d) 7. (a) Message flow itialization phase competition; (b,c) 7. (a) Message flow itialization phase competition; (b,c) flow at 7. (a) stage Message flow itialization competition; phase (d) flow after competition; (b,c) competition. flow at stage competition; (d) flow after competition. flow at stage competition; (d) flow after competition Message Flow durg Path Selection Message Flow durg Path Selection After g, s beg to build a connected path. When a s dtance to sk After g, less than dhot, such s as beg to build a connected path. When a s dtance to p2 6c, sends a request to sk sk less than dhot, such as p2, sk replies with a confirmation 6c, sends to a ; request process to sk, 8a. When sk a s replies dtance with to a confirmation sk greater to than ; process dhot re are or p 6 p 6

14 Sensors 2016, 16, Message Flow durg Path Selection After g, s beg to build a connected path. When a s dtance to sk less than d hot, such as 6c, sends a request to sk, sk replies with a confirmation to ; process 8a. When a s dtance to sk greater than d hot re are or s with its R c _max, such as p 6 6c, it has a exchange with a, which 8b. In a similar way, when a has no neighbors with its R c _max, such as p 1 6c, it has a exchange with a non- ; flow between Sensors 2016, m 16, 98 8c. Sensors 2016, 16, 98 p 6 p 6 R R RC RC (a) (a) R R RC RC (b) Sk Clust p (b) R 1 er- R RC Clust p 1 RC (c) er- RC: replay confirmation R: Request (c) next-hop RC: replay confirmation R: Request next-hop 8. Message 8. Message flow flow durg durg path path selection. (a) (a) Message flow flow with with Sk Sk ; (b) (b) Message flow with with 8. ; Message (c) ; Message flow durg (c) Message flowpath with flow selection. non- with non- (a) Message.. flow with Sk ; (b) Message flow with ; (c) Message flow with non Message Message Flow Flow durg durg In-Cluster In-Cluster ode ode Position Position Adjustment Message Flow durg In-Cluster ode Position Adjustment In th process, flow between s takes place three sub-processes. First, a In th process, flow between s takes place three sub-processes. First, a In th process, receives MCD from flow its between next-hop s takes place begs to three calculate sub-processes. position First, its a - s, receives such receives as M p2 CD from its 6c. Second, after position begs calculation to calculate - position s, its MCD from its next-hop begs to calculate position its - - sends s, s, such to such its as- pas 2 p2 s. 6c. 6c. Last, Second, it sends after MCD to position one calculation its last-hop s - - if it cannot s, s, be it it sends sends adjusted. Orwe, to its to its - sends s. MAD Last, to it its sends next-hop MCD M CD. to toone one The its its last-hop flows s between s if it if cannot m it cannot are be be adjusted. adjusted. respectively Orwe, 9a d. sends MMAD to its next-hop. The The flows flows between between m m are are respectively respectively 9a d. 9a d. Sk MCD(Calculate divg depth) Sk MCD(Calculate divg (a) depth) Depth, Clust (a) basic er- Depth, (b) Clust basic er- MCD(Calculate (b) divg depth) p 6 MCD(Calculate divg (c) depth) p 6 MAD(Been (c) adjusted) Sk MAD(Been adjusted) (d) Sk (d) 9. Message flow durg position adjustment. (a) Initialization phase position adjustment; (b) 9. Phase Message 9. Message position flow flow durg adjustment; durg position (c,d) adjustment. End position (a) (a) adjustment. Initialization phase position adjustment; (b) Phase(b) Phase position position adjustment; adjustment; (c,d) End (c,d) End position position adjustment Complexity Analys 4.2. Complexity Analys 4.2. Complexity The Analys time complexity URSA are evaluated th section. The time complexity URSA are evaluated th section. The Message Complexity time complexity URSA are evaluated th section Message Complexity Message The total Complexity number sent s considered to determe complexity. The The total complexity number cludes sent three s parts, accordg considered to to determe analys complexity. flow between The The s total complexity Section number 4.1. cludes sent s three parts, accordg considered to to determe analys flow complexity. between The s complexity s Section are 4.1. cludes scattered three on parts, water accordg surface romly to analys uniformly. At begng flow between s Section uneven 4.1. s g, are scattered p s on become water provional surface romly s, uniformly. each At m begng broadcasts its uneven own status g, p ( s ID, become d(pi,sk), provional R(pi)), where p formulated s, each Equation m (17) broadcasts accordg its to own Equation status (8). ( ID, d(pi,sk), R(pi)), where p formulated Equation (17) accordg to Equation (8). 2 dhot Th1 3T h2 4Th 3 p 2 T d h3 (17) hot length Th1 3width Th 2 4Th 3 p Th3 (17) length width Each m n makes a decion by broadcastg MCH to act as fal or Sk

15 Sensors 2016, 16, s are scattered on water surface romly uniformly. At begng uneven g, p s become provional s, each m broadcasts its own status ( ID, d(p i,sk), R(p i )), where p formulated Equation (17) accordg to Equation (8). p πd2 hot pt h1 ` 3T h2 4T h3 q length ˆ width ` T h3 (17) Each m n makes a decion by broadcastg M CH to act as fal or M GC to exit competition. Assumg that re are K c s, y send K c elect s, n ( K c ) non- s send ( K c ) jo s. Thus, process uneven g, total number formulated as: 2 p ` K c ` p K c q 2 p ` (18) Durg process path selection, each sends a request to its next-hop, n, a correspondg reply sent to it. Thus, total number s 2K c process. In phrase - s position adjustment, each calculates positions - s, when it receives M CD from its next-hop. After positions s its are adjusted, it sends M AD to its next-hop sooner or later. Therefore, re must be two s transmitted on lk between a its next-hop. What more, each has only one next-hop because hybrid radius path-selection method. Therefore, network has K c s, a sk has K c lks; number transmitted on se lks 2K c. In addition, each respectively sends a correspondg adjustment to its - s, after calculatg depths s its. Supposg that i-th has i - s, number s sub-process formulated as: Fally, total number whole process 2K c +. Consequently, total number URSA s formulated as: ÿk c i 1 i (19) 2 p ` ` 2K c ` 2K c ` 2 p ` 2 ` 4K c ă 8 (20) Thus, complexity URSA O() Time Complexity At begng uneven g, each synchronously produces a rom number from zero to one estimates itself to be a provional or not, n, each provional s synchronously broadcasts its status. The time complexity se operations constant. In competition phrase, a provional may become a fal, which needs to compare dtance to sk with all its competition neighbor s or exit competition. For convenience expression on time complexity, number competition neighbor s a provional expressed by average number competition neighbor s network, signed as a. Consequently, time complexity operation that a provional becomes a fal a, time complexity operation that it exits competition constant. However, worst situation phrase that each m judges itself to be a fal or not one-by-one way, time complexity situation a ˆ K c + p K c + 2. At end