Energy-efficient scheduling and hybrid communication architecture for underwater littoral surveillance

Transcription

1 Computer Communictions xxx (26) xxx xxx Energy-efficient scheduling nd hybrid communiction rchitecture for underwter littorl surveillnce Mihel Crdei * Deprtment of Computer Science nd Engineering, Florid Atlntic University, Boc Rton, FL 33431, USA Received 9 Jnury 26; ccepted 9 Jnury 26 Abstrct There exists high demnd for relible, high cpcity underwter coustic networks to llow efficient dt gthering nd informtion exchnge. This is evidenced by significnt reserch in overcoming the limittions of the shllow wter coustic chnnel, such s low bndwidth, highly vrying multipth effects nd lrge propgtion delys. This pper proposes n energy-efficient scheduling mechnism for sclble network topology with hybrid RF-coustic communiction rchitecture, designed for underwter littorl surveillnce pplictions. In combintion with sclble buoy positioning scheme tht ssures AUV underwter synchroniztion nd position computtion, we propose Time Division Multiple Access scheduling technique for underwter communiction tht chieves energy-efficient, collision-free dt exchnge on the low dt rte coustic chnnel. We mesure the performnce of the communiction rchitecture by extensive simultions using the Opnet network simultor. Ó 26 Elsevier B.V. All rights reserved. Keywords: Ad hoc wireless networks; Network topology; Energy-efficient scheduling; Communiction rchitecture; Autonomous underwter vehicles; Underwter littorl surveillnce 1. Introduction Recent dvnces in the shllow-wter underse monitoring, surveillnce nd explortion demnd using the underwter coustic communiction chnnel s the primry medium for dt collection nd informtion exchnge. The communiction network usully includes utonomous underwter vehicles (AUVs) interconnected by coustic modems nd one or more surfce gtewys which provides links to commnd nd control center, which cn further be connected to bckbone network, such s the Internet. Surfce communiction cn be done using RF or stellite links. Applictions of deploying AUVs in shllow-wter includes underse surveillnce in littorl wters, ocenogrphic dt gthering, environmentl monitoring, nd costl defense. One of the mjor chllenges in underwter coustic networking is the development of networking protocols to * Tel.: ; fx: E-mil ddress: mihel@cse.fu.edu. overcome hrsh communiction conditions. Physicl lyer considertions include limited bndwidth, extreme propgtion delys, signl bsorption, time-vrying multipth with severe intersymbol interference nd lrge Doppler shifts [13]. These physicl lyer limittions dversely impct throughput, ltency, nd cpcity. Since the coustic network components re btterypowered, power efficiency is n importnt chrcteristic of the underwter communiction protocols. Designing energy-efficient scheduling nd communiction mechnisms hve n importnt role in prolonging the network opertionl lifetime. There exist lrge volume of reserch ddressing the difficulties of underwter coustic networks nd proposing solutions t different lyers of the network rchitecture. In [13,15], uthors present comprehensive survey of existing network technology t the physicl, dt link nd network lyers nd their pplicbility to underwter coustic chnnels. A cluster-bsed communiction protocol is proposed in [11]. Adjcent underwter vehicles re grouped into clus /$ - see front mtter Ó 26 Elsevier B.V. All rights reserved. doi:1.116/j.comcom

2 2 M. Crdei / Computer Communictions xxx (26) xxx xxx ters in which in-cluster communiction is chieved through TDMA nd inter-cluster communiction uses CDMA. The focus of this pper is on cluster mintennce, tht needs to ccommodte dynmic topology with mobile vehicles. The Deployble Autonomous Distributed Systems (DADS) [1] re motivted by requirement for wide-re underse surveillnce in littorl wters. The network connects the remote sensor pltforms through gtewy (e.g., se-surfce buoy) to distnt mster node cross stellite links. Acoustic dt is trnsmitted over multihop pths, using hlf-duplex CDMA links between discrete modem pirs nd three-wy hndshke protocol for dt trnsfer between ech pir of nodes. The pper lso describes series of ocen experiments nd mesurements. The work in [18] proposes networking protocol for underwter coustic networks of fixed or mobile nodes. The uthors propose centrlized topology control scheme by periodic topology discovery probes. The routing protocol proposed is centrlized, where the mster node determines ll routes through the network. CDMA technique is used for multiple ccess to the communiction medi. In this pper, we design shllow wter energy-efficient scheduling mechnism nd communiction rchitecture for littorl surveillnce ppliction. Our model includes (1) number of AUVs deployed for shllow-wter underse surveillnce, (2) number of se-surfce buoys tht ct s gtewy between the underwter coustic network (ACnet) nd the RF network (RFnet), nd (3) centrl commnd nd control sttion on ship positioned in close proximity to the surveillnce re. The reminder of this pper is structured s follows. Section 2 presents the networking model with prmeters considered for ACnet nd RFnet. In Section 3, we describe the littorl surveillnce scenrio nd problem definition. Next, we present the network rchitecture design in Section 4. We continue with our energy-efficient scheduling mechnism. Buoys rrngement mechnism is presented in Section 5, followed by the TDMA slot lloction lgorithm in Section 6. Section 7 presents performnce nlysis for simultion results nd Section 8 concludes our pper. 2. Hybrid networking model Underwter communiction is severely limited by the physicl lyer chrcteristics of the communiction medium. Usully, coustic networks in shllow wter support dt rte of 1 1 bps [18], lthough higher dt rtes of 16 kbps re reported in [12]. The vilble bndwidth for n coustic network is typiclly round 15 khz, nd it hs to be shred mong ll coustic nodes. Three mjor multiple ccess methods re Time Division Multiple Access (TDMA), Frequency Division Multiple ccess (FDMA), nd Code Division Multiple Access (CDMA) [17]. The work in [13] compres these methods from the point of view of shllow wter coustic communiction network. CDMA nd spred-spectrum signling pper to be promising multiple ccess technique. Another key fctor is the coustic signl propgtion speed of pproximtely 15 m/s, five orders of mgnitude lower thn tht of rdio chnnel. The lrge propgtion delys directly ffect the chievble throughput. Also, the signl rnge of n AUV vries from few kilometers up to 9 km [4]. The ppliction we re ddressing in this pper is underwter littorl surveillnce. The environment we consider is shllow wter, with depth less thn 5 feet. The hybrid network topology we consider is illustrted in Fig. 1 nd its components re described next. The rchitecture elements in our model re AUVs, buoys, nd centrl commnd nd control sttion, hosted on ship. Buoys nd the centrl sttion re considered fixed pltforms nd the AUVs re mobile sensor pltforms. In this pper, we use chrcteristics of the wireless equipment vilble t the Florid Atlntic University (FAU), the Deprtment of Ocen Engineering [1]. AUVs nd buoys cn communicte wireless to synchronize nd to shre informtion. Ech AUV nd buoy is equipped with FAU Dul Purpose Acoustic Modem (FAU-DPAM) [1,8]. A DPAM hs two trnsceivers: the first trnsceiver hs n RF rdio with 2.4 GHz trnsmitter with omnidirectionl ntenn for the surfce wireless communiction, hving 3 km rnge. Fig. 1. Hybrid RF-Acoustic Network Components.

3 M. Crdei / Computer Communictions xxx (26) xxx xxx 3 We ssume dt rte of 1 Mbps. Higher dt rte COTS products re vilble [14]. The ship hs only RF connectivity. the second trnsceiver is using n coustic chnnel in the khz bnd for underwter coustic communiction; the communiction rnge is bout 3 km with dt rte of 3 bps using Direct Sequence Spred Spectrum (DSSS) with Multiple Frequency-Shift Keying (MFSK) or it cn be bout 5 m with dt rte of 15 kbps by using DSSS with Phse-Shift Keying (PSK) [17]. In this pper, we ssume communiction rnge of 3 km nd dt rte of 3 bps. The AUVs re lmost ll the time moving with speed less thn 3 m/s. Their min tsk in this model is to monitor the costl ocen environment nd detect mines nd other suspect objects. Ech AUV is equipped with Side Scn Sonr HF Acoustic chirp [2] t 5 khz, with sensing rnge of bout 1.5 miles. Every AUV is lso equipped with cmer. When mine or nother potentilly dngerous object is detected, the AUV tkes one or more pictures (of bout 25 kbytes ech) tht need to be sent immeditely to the ship. Additionlly, n AUV is lso equipped with Globl Positioning System (GPS) receiver used for periodic time synchroniztion nd positioning when the AUV surfces. Fixed buoys provide communiction support for littorl AUV missions. In our model, buoys re deterministiclly plced t the beginning of the mission by specilized submrines. Buoy deployment positioning is detiled in Section 5. Buoys re equipped with GPS, llowing them to precisely determine time nd their loction. Buoys periodiclly brodcst their loction on the coustic network, llowing AUVs within trnsmission rnge to tringulte their own loction while being submerged. Both AUVs nd buoys re bttery-powered, nd therefore preserving energy resources by scheduling nd efficient communiction will increse the network lifetime. Beside AUVs nd buoys, the topology contins centrl commnd nd control sttion on ship. In our model, we ssume the ship is positioned within RF communiction rnge of t lest one buoy. Our rchitecture ssumes the ship is moving t low speed, or is sttionry. The hybrid wireless network consists of two d hoc wireless networks. The RF wireless network (RFnet) runs bove the ocen surfce, nd is formed by buoys, the ship nd the AUVs (while t surfce). The RFnet hs dynmic topology t the edges, becuse every AUV joins this network only when it reches the surfce nd must communicte t high speed. The RF network hs dt rte of 1 Mbps in the 2.4 GHz frequency bnd. Trnsceivers for RFnet re vilble for long rnge outdoors communiction. For exmple, MikroTik Wireless LAN USB dpter [14] uses 2.4 GHz DSSS rdio trnsmission, for dt rte of 11 Mbps, nd hs rnge of up to 5 km for outside environments. The trnsmit power is 32 mw. The second wireless network is the underwter ACoustic network (ACnet). The nodes in the ACnet re buoys nd AUVs nd they communicte using the coustic trnsceiver on the FAU-DPAM modem. As we discussed in the previous prgrphs, the communiction in the coustic network is severely limited by the physicl properties of the coustic chnnel in shllow wter. The interfce between the RFnet nd ACnet cn be ccomplished in two wys. First, buoys re prt of both networks, bsed on their dul modem, therefore they re ble to trnsfer informtion between the two networks. Second, the AUVs cn surfce nd join the RFnet. 3. Problem definition nd surveillnce scenrio In this pper, we ddress the underwter littorl surveillnce ppliction. A number of AUVs re deployed nd re using their sonr sensing cpbilities to detect mines or other relted trgets. Once suspicious object is detected, trget informtion needs to be reported immeditely to the ship, cross the wireless hybrid network. Our gol in this pper is to design n energy-efficient scheduling mechnism nd communiction rchitecture solution for underwter littorl surveillnce pplictions with the following requirements: (1) ccommodte the hrsh underwter coustic communiction prmeters (2) to relibly trnsmit to the ship, with high priority, messges from AUVs bout detected trgets (3) to periodiclly nd relibly send updtes with AUV sttus nd loction (4) to periodiclly trnsmit updtes to the ship on buoy sttus, nd (5) to relibly trnsmit commnds nd/or queries from ship to AUVs or buoys. With these requirements, we describe littorl surveillnce mission scenrio consisting of the following phses: Buoys nd AUVs deployment. This phse occurs t the beginning of the mission. The buoys nd AUVs re deployed by specilized submrine or by ship. We ssume the buoys re deterministiclly plced, s specified in Section 5. We lso ssume every AUV nd buoy hs ssigned unique ID. After deployment, the buoys nd AUVs strt the network initiliztion step. Network initiliztion. During the initiliztion phse, buoys nd the ship orgnize into n d hoc RF wireless network (RFnet) nd strt the routing protocol initiliztion. This opertion consists in exchnging short control messges nd constructing the routing tbles, in ccordnce with the routing protocol employed. Once n AUV is deployed, it goes to the surfce nd sends n initiliztion messge to the ship over the RFnet. This messge contins the AUV ID nd its current loction. Bsed on this informtion nd using our scheduling lgorithms, the ship replies with messge contining coordintes of the surveillnce region (where this AUV will perform the sensing tsk) nd TDMA slot. This TDMA slot will be used by the AUV for communiction in the ACnet. Division of

4 4 M. Crdei / Computer Communictions xxx (26) xxx xxx the surveillnce re in surveillnce regions is described in Section 5. TDMA slot lloction is discussed in Section 6. Idle network opertion. This phse describes the network opertion between trget detection events. The min tsk of the RFnet is to mintin the routing informtion updted, by periodiclly exchnging control informtion in ccordnce with the routing protocol used. Every buoy is responsible for periodiclly sending sttus report messges to the ship, contining informtion bout its stte (e.g., OK/CriticlCondition) nd list with AUVs in its rnge. We will refer to this messge s the sttusupdte messge. Next, we describe the opertions performed in the ACnet. Every AUV nd buoy will periodiclly send short messge with nvigtion control informtion. We refer to these messges s buoy becons nd AUV becons, respectively. A buoy becon contins the buoy ID, GPS-bsed loction nd time informtion, nd other ppliction specific informtion. For exmple buoy becon my encode short commnd sent from the ship to n AUV. An AUV becon contins the AUV ID, flight time, loction, speed, bering, nd one or more fields used for ACK nd dt exchnge. Ech AUV listens to other buoy becons nd AUV becons. Bsed on our design, every AUV moves into surveillnce region with the property ny point is within underwter coustic trnsmission rnge of t lest three buoys. Using loction nd time informtion from t lest three buoys, the AUV is ble to synchronize its clock nd compute its current loction through tringultion [5] while submerged. Informtion from other AUV becons is used for collision voidnce nd other nvigtion purposes. Every time n AUV gets to the surfce, it will use its GPS for precise time nd loction lignment. The system is lso chrcterized by GPS-updte-time ttribute. At most every GPS-updte-time, n AUV surfce to get synchronized. Every buoy listens for AUV becons nd updtes its AUV dtbse informtion. The min opertion performed by AUVs is sensing. Every AUV is equipped with Side Scn Sonr HF Acoustic chirp. The AUVs nvigte nd continuously monitor their environment. Trget detection. This phse strts when n AUV detects mine or suspicious trget. Ech AUV is equipped with digitl cmer. When trget is detected, the AUV cptures one or more photogrphs tht need to be sent with high priority to the ship. Becuse of the limittions in dt rte nd signl dely in the ACnet, it is not fesible to hve AUVs send lrge dt messges (e.g., pictures) over long distnces on the underwter network. The solution we dopt is to hve n AUV tht detects trget go to the surfce, join the RFnet nd strt messge exchnge with the ship. In our communiction protocol, the AUV sends in the fist messge one picture to the ship nd then wits for n cknowledgment, query or commnd from the ship. While witing for the reply, the AUV performs GPS loction nd time updte. If n cknowledgment is received, the AUV dives nd continues the surveillnce process. The AUV cn lso receive query from the ship, sking for more pictures (lredy vilble t the AUV), cse in which the AUV strts trnsmitting the required informtion nd wits for new ship reply. Besides these two replies, the ship my commnd the AUV to go to specific loction within its surveillnce region nd monitor the environment. In this cse, the AUV dives, strts the sonr, sets the new wypoint nd begins its new tsk. New sensing tsk. The ship cn explicitly sk n AUV to nvigte t specific loction in its surveillnce region nd collect dt. The messge is sent by the ship cross the RFnet to the buoys tht re within this AUV s communiction rnge. These buoys will then trnsmit this commnd on the ACnet, piggybcked in buoy becon. Mission termintion. This opertion is ccomplished by hving the ship sending specil Termintion control pcket to every AUV. Upon receiving this messge, n AUV performs the termintion procedure, nvigting towrd specific pick-up loction. 4. Hybrid network rchitecture design In this section, we describe the networking rchitecture for the ACnet nd RFnet. In n underwter coustic environment, the communiction mechnism needs to consider poor conditions in terms of ltency nd dt rte. We design the ACnet rchitecture to include only Physicl, Dt Link, nd Appliction Lyers. Our model does not require network protocol, becuse the only communiction in ACnet is AUV nd buoy becon brodcst. This is n one-hop informtion exchnge, where nodes do not forwrd the messge they receive. The Physicl Lyer uses DSSS with MFSK with dt rte of 3 bps nd communiction rnge R = 3 km, ccording with the chrcteristics of the FAU-DPAM modem, coustic trnsceiver (see Section 2). The multiple ccess scheduling method we dopt for the underwter communiction is TDMA. TDMA voids collisions by llocting time slot to every AUV nd buoy. It is n energy-efficient scheduling mechnism since it voids dditionl control messges exchnge. For exmple, TDMA scheduling mechnism voids the hndshke overhed of wireless MAC protocols such s Crrier Sense Multiple Access with Collision Avoidnce (CSMA-CA), nd Multiple Access with Collision Avoidnce-derived protocols like MACAW (MACA for Wireless) nd IEEE [16] protocols. Actully, MAC trnsmission protocol involving lrge number of messge exchnge is lso not fesible becuse of the long propgtion delys in coustic networks.

5 M. Crdei / Computer Communictions xxx (26) xxx xxx 5 Bsed on the buoy deployment mechnism (see Section 5), five TDMA slots from the TDMA frme will be ssigned to the buoys, s represented in Fig. 3 nd ll other will be ssigned to the AUVs. The scheduling using the TDMA slot lloction mechnism is further detiled in Section 6. AUVs nd buoys use the TDMA slots for sending short becons in the ACnet nd not for lrge dt trnsmissions. For exmple, if we design becon with 32 bytes, then we cn hve TDMA slot of 3 s. This will suffice for.85 s becon trnsmission delys nd 2 s propgtion dely t 3 km distnce. The RFnet network is used for trnsmitting long messges between the AUVs nd ship, for sttus-updte messges nd to issue commnds from the ship to the AUVs. A ship commnd ddressed to submerged AUV will be sent to the buoys in the communiction rnge of tht AUV. This is esy to implement, bsed on informtion from the sttus-updte messges received by the ship from buoys. The RFnet rchitecture is typicl to n IP d hoc wireless networks nd includes the Physicl nd Dt Link lyers, nd the TCP/IP Network nd Trnsport lyers. At the Physicl Lyer, we ssume the following ttributes: rdio frequency 2.4 GHz, DSSS signl spreding technique, dt rte of 1 Mbps nd communiction rnge of 3 km. The RF communiction rnge should be greter or equl to the coustic communiction rnge in order to hve ny AUV t the surfce within RF trnsmission rnge of t lest one buoy. The MAC protocol used in the RFNet is IEEE Distributed Coordintion Function (DCF). At the Network Lyer, we use rective routing protocol for d hoc wireless networks, such s Dynmic Source Routing (DSR) or Ad Hoc On-Demnd Distnce Vector (AODV). In our simultions we use DSR for its good performnces in terms of relibility, routing overhed nd pth optimlity [6,7]. When n AUV joins the RFnet for informtion exchnge it lso runs the selected d hoc routing protocol but is not involved in dt forwrding on behlf of other nodes. It is only lef in the network topology. At the Trnsport Lyer, we use the TCP protocol [16] for its relible communiction nd congestion voidnce mechnism. It is well-known tht TCP performs poorly in multihop wireless networks [3], due to its inbility to differentite between retrnsmissions due to congestion versus pcket loss. In our ppliction scenrio AUVs find trgets infrequently nd ll high dt rte trffic is sent towrds the ship node, thus congestion my occur. 5. Buoy positioning The positioning of the buoys plys n importnt role in our scheduling mechnism. The fctors we considered in designing mechnism for computing the buoy loctions re the following: Mission specifics. The mission considered in this pper is littorl surveillnce, where AUVs re moving long the costline. AUV loction. AUVs need to be ble to tringulte their loction bsed on the GPS informtion included in the periodic buoy becons. At ny time ech AUV should be within trnsmission rnge of t lest three buoys. Minimized number of buoy slots. The underwter communiction protocol proposed in this pper is TDMA, nd, becuse the frme period is limited (see Section 6), the number of slots llocted to the buoys should be minimized. Mximized surveillnce re. The surveillnce re where AUVs nvigte should be mximized. Note tht ny point within the nvigtion re should be within the communiction rnge of t lest three buoys for loction identifiction. Bsed on these considertions, we rrnge the buoys in two rows long the costline in order to provide rectngulr surveillnce re (see Fig. 2) with the property tht ny point within this re is within communiction rnge of t lest three buoys. Therefore, ny AUV nvigting within this rectngle receives periodic becons from t lest three buoys, nd will be ble to periodiclly updte its loc- Fig. 2. Buoy positioning for n = 8 buoys.

6 6 M. Crdei / Computer Communictions xxx (26) xxx xxx tion through tringultion. The distnce between two buoys in the sme row is equl to R, where R is the underwter communiction rnge of the buoys nd AUVs (e.g., R is set up to 3 km). Next, we will determine the reltive loction of the two rows in order to mximize the surveillnce re. Let us consider the first row fixed nd the Crtesin coordinte system s in Fig. 2. We neglect in this discussion the verticl coordinte, becuse the mission is performed in shllow wter, with depth less thn 5 feet, much smller thn the vlues in the other two horizontl coordintes. We consider n buoys B 1, B 2, B 3,..., B n deployed s in Fig. 2. We orgnize them in two rows, with B 1, B 3,... in the first row nd B 2, B 4,... in the second row. Next, we compute the loction of the second row (e.g., determine d x nd d y ), in order to mximize the surveillnce rectngle re. In the picture, we hve represented the two buoy rows nd their communiction disk, inside circle with rdius R. In the Fig. 2, we cn observe tht the points U 1, U 1, U 2, U 2,...nd L 1, L 1, L 2, L 2,...estblish the upper nd lower rectngle edges. The points right bove U 1, U 1, U 2, U 2,... or below L 1, L 1, L 2, L 2,... re not within communiction rnge of three buoys. The rectngle width cn be expressed s W ¼ minfy U 2 y L2 ; y U y 2 L g, where x A nd y A re the 2 x nd y coordintes of the point A. Considering d y fixed, bsic geometric computtion indictes tht W is mximized when d x = R/2, cse when both U 1, U 1, U 2, U 2 ;... nd L 1, L 2, L 2, L 3,... re colliner. The left edge of the rectngle is determined by the intersection point (e.g., U 1 ) of the circle with rdius R centered t the third buoy from left (e.g., B 3 ) with the upper edge, nd the right edge of the rectngle is determined by the intersection point (e.g., L 3 ) of the circle with rdius R centered t the third buoy from right (e.g., B 6 ) with the lower edge. Note tht there will be other points in the rectngle s vicinity tht will be covered by t lest three buoys, but our gol is to determine rectngulr shpe with this property. p Next we wnt to determine d y 2½; R ffiffi 3 =2Š such tht to mximize p the surveillnce rectngle re, Are = W L. d y ¼ R ffiffi 3 =2 corresponds to the p cse when B 1 B 2 = B 2 B 3 = B 1 B 3 = R nd W ¼ R ffiffi p 3 =2. Cses d y > R ffiffi p 3 =2 re not considered, s they produce W < R ffiffi 3 =2. Using the geometric computtion presented rffiffiffiffiffiffiffiffiffiffiffiffi in Appendix A, we obtin tht W ðd y Þ¼ 3R 7R 2 4y 2 d nd 2 9R 2 þ4y 2 d rffiffiffiffiffiffiffiffiffiffiffiffi 7R Lðn; d y Þ¼ðn 2ÞR=2 d 2 4d 2 y y for ny n P 3. Tking 9R 2 þ4d 2 y dðw ðdyþþ dðd yþ the derivtive, we obtin 6 nd dðlðn;dyþþ dðd yþ <, when p d y 2½; R ffiffiffi 3 =2Š. Therefore, W (dy ) nd L n (d y ) re decresing functions of d y, chieving the mximum vlue when d y =. An importnt gol of deploying buoys is to brodcst GPS informtion. In order for the AUVs to perform the tringultion lgorithm, the buoys cnnot be colliner, therefore d y needs to be greter thn specific threshold. We ssume d y = d, where d is the minimum vlue required in the tringultion lgorithm. Note tht this buoy orgniztion is sclble. Additionl buoys cn be deployed to increse the surveillnce length. Any dditionl buoy will increse the length with R/2. Also, by using this rrngement, only five TDMA slots will suffice for the buoys coustic communiction. TDMA slot lloction is presented in Section 6. Another observtion is tht buoys re nchored t the ocen floor, therefore their loction on the ocen surfce is not fixed, they re continuously moving round the nchor point. Still this distnce is very smll compred to the distnces between buoys nd it is therefore neglected in this pper. 6. TDMA slot lloction In this pper, we propose TDMA scheduling mechnism both for energy-efficiency considertions, s well s becuse of the coustic environment constrints, s discussed previously. TDMA frme period is limited by the frequency with which the AUV nd buoys need to send becons. The becons re criticl for AUV synchroniztion, positioning, nd nvigtion s well for sending ccurte sttus updte informtion. On the other hnd, the TDMA slot size is ffected by the low dt rte nd long propgtion dely in the coustic network. Therefore, the gol of the TDMA slot lloction mechnism is to use minimum number of TDMA slots in frme intervl. In our scenrio, the first five slots 1 5 re llocted to the buoys. Considering the numbering in Fig. 3, the buoy B i, i = 1,2,..., n will be ssigned the slot i In this wy, ny AUV in the surveillnce re will be within communiction rnge of t most one buoy trnsmitting in the sme slot. This is becuse the communiction rnge of ny two buoys ssigned to the sme slot (e.g., B 1 nd B 6 ) do not intersect. Let us ssume tht fter llocting the first five slots to the buoys, we re left with s TDMA slots to be llocted to the AUVs. Consider the number of AUVs to be deployed is m nd the length of the surveillnce rectngle is L. Next, we present mechnism for llocting slots to the AUVs, using sptil slot reuse. This mechnism is performed only once, during the initiliztion phse. The slot lloction lgorithm is run by the ship. The following mechnism cn be pplied when L/ (r +1)>R, where r is the number of regions computed by Algorithm 1. If this reltion does not hold, then we need to djust the L, m, nd s prmeters correspondingly, e.g., either increse L or the number of slots s or decrese m, the number of AUVs. Next we present two lgorithms. Algorithm 1 hs s input m, s, L nd divides the surveillnce rectngle in

7 M. Crdei / Computer Communictions xxx (26) xxx xxx 7 Fig. 3. Surveillnce regions for n = 15 buoys nd time slot lloction when m = 1 nd s =6. regions, ssocites TDMA slots to every region nd identifies every buoy to which region it belongs bsed on its loction. Any slot ssocited to region is lter ssigned by the ship to n AUV to use for becons while performing the surveillnce tsk within tht region. Algorithm 2 ssigns communiction slot to n AUV. This lgorithm is executed by the ship s result of request from the AUV sent during the initiliztion phse, when the AUV is t the ocen surfce. Let us note with S the set of TDMA slots for AUVs, e.g., for exmple in Fig. 3, S ={s 6,s 7,...,s 11 } nd S =s, where S represents the crdinlity of the set S. Algorithm 1 (m, s, L). Input: m, s, L Output: divide surveillnce rectngle in regions, ssign TDMA slots to ll regions Step 1. If m 6 s, then the surveillnce rectngle forms one region R 1. Assign ll s slots to R 1. All buoys will lso be prt of R 1. Exit. Step 2. Compute the number of regions r First, we divide the slots in S in three sets S 1, S 2, S 3, S 1 + S 2 + S 3 =s: if (s == 3k) where k 2 N, then S 1 = S 2 = S 3 =k else if (s == (3k + 1)) then S 1 =k + 1 nd S 2 = S 3 = k else if (s == (3k + 2)) then S 1 = S 2 =k + 1 nd S 3 =k Let S 1 contin the first S 1 slots from S, S 2 the next S 2 slots, nd S 3 the lst S 3 slots. Next, compute the number of regions r s follows: r =3ºm/sß if ((m s) >( S 1 + S 2 )) then r = r +3 else if ((m s) > S 1 ) then r = r +2 else if ((m s) > ) then r = r +1 Step 3. Divide the surveillnce rectngle into r equl regions This step divides the surveillnce rectngle into r + 1 rectngles, with length d = L/(r + 1), s illustrted in Fig. 3. Note tht L/(r +1)>R. We define r regions s the following rectngles: R 1 ¼ R 1 R 3 R 3 R 1 ; R 2 ¼ R 2 R 4 R 4 R 2 ;...; R r ¼ R r R rþ2 R rþ2 R r. Step 4. Assign slots to every region R i, for i =1,...,r for (i =1,...,r) { if ((i 3) == 1) then ssign to R i the slots in the set S 1. else if ((i 3) == 2), then ssign to R i the slots in the set S 2. else if ((i 3) == ), then ssign to R i the slots in the set S 3. } Step 5. Every buoy belongs to one, two or three regions depending on its loction. for (i =1,...,n, j =1,...,r){ if (x Bi 2 [x Rj,x Rj+2 ]) then B i belongs to the region R j, where R j ¼ R j R jþ2 R jþ2 R j } Algorithm 2. Input: AUV i Output: Allocte TDMA slot to AUV i Bsed on the AUV i loction, determine the closest region R j tht hs slot u vilble. If two such regions re vilble, choose the one with smllest index. Mrk the slot u in R j s busy nd ssign slot u to AUV i. AUV i will belong to R j, this mens tht AUV i will perform the surveillnce opertion within R j nd will trnsmit its becon during the slot u in the ACnet. Theorem 1. Algorithm 1 provides enough slots to ccommodte ll m AUVs trnsmissions. Any buoy nd ny AUV nvigting in region R i will receive collision-free messges from ny AUV tht hs been ssigned slot from R i s result of Algorithm 2. Proof. First, we prove tht the number of slots llocted by the Algorithm 1 in ll r regions is greter or equl with m. We distinguish four cses:

8 8 M. Crdei / Computer Communictions xxx (26) xxx xxx 1. (m s) ==. Then there re r = 3m/s regions nd number of slots is (m/s)( S 1 + S 2 + S 3 ) = m. 2. (m s)>( S 1 + S 2 ). In this cse we hve r = ºm/sß +3 regions nd the number of slots llocted to ll regions is (ºm/sß + 1)( S 1 + S 2 + S 3 ) = (ºm/sß +1)s > m. 3. ( S 1 + S 2 ) P (m s)> S 1. In this cse we hve r = ºm/sß + 2 regions nd the number of slots is (ºm/sß) ( S 1 + S 2 + S 3 ) + S 1 + S 2 =(ºm/sß)s + S 1 + S 2 >m. 4. js 1 P (m s) >. In this cse we hve r = ºm/sß +1 regions nd the number of slots is (ºm/sß)( S 1 + S 2 + S 3 ) + S 1 = (ºm/sß)s + S 1 >m. If n AUV hs been ssigned slot u from region R i, then it will nvigte nd perform surveillnce within tht region. Bsed on the wy we ssign slots to regions (see Algorithm 1, Step 4), the closest regions with the sme slots re R i nd R iþ3, for i =1,...,r 3. The minimum distnce between R i nd R iþ3 is d nd becuse d > R, ny node (AUV or buoy) inside R i will not receive ny trnsmission from node inside R iþ3. h Let us now show the slot lloction mechnism on the exmple illustrted in Fig. 3. We know L, m = 1, S = {6,7,8,9,1,11}, s = 6. Then S 1 = {6,7}, S 2 = {8, 9} nd S 3 = {1,11} nd there re five regions: R 1 ¼ R 1 R 3 R 3 R 1 ; R 2 ¼ R 2 R 4 R 4 R 2 ; R 3 ¼ R 3 R 5 R 5 R 3 ; R 4 ¼ R 4 R 6 R 6 R 4 ; nd R 5 ¼ R 5 R 7 R 7 R 5. Time slots in the set S 1 re ssigned to R 1 nd R 4, time slots in S 2 re ssigned to R 2 nd R 5 nd time slots in S 3 re ssigned to R 3. Assuming the ten AUVs re deployed uniformly long the rectngle, possible slot ssignment is s follows: R 1 : AUV 1 on slot 6 nd AUV 2 on slot 7, R 2 : AUV 3 on slot 8 nd AUV 4 on slot 9, R 3 : AUV 5 on slot 1 nd AUV 6 on slot 11, R 4 : AUV 7 on slot 6 nd AUV 8 on slot 7, R 5 : AUV 9 on slot 8 nd AUV 1 on slot 9. Advntges of using this slot lloction mechnism include: (1) energy-efficient mechnism with low overhed; (2) it uses reduced number of time slots s; (3) it proposes uniform AUV distribution, by orgnizing the surveillnce re in smller regions; (4) it is sclble. More buoys nd AUVs cn be deployed following the sme pttern, without modifying the existing orgniztion. 7. Simultion results In this section, we nlyze the performnce of our protocol through simultions. We implement the communiction protocol with the Opnet network simultor [9]. Next, we present the vlues of the prmeters considered in our testing, followed by the performnce metrics nd simultion results. For testing, we ssume scenrio with n=15 buoys, positioned ccording to Algorithm 1 (see Fig. 3). This will result in surveillnce re with length L = m for d y = 1 m. The RF communiction rnge is R =3km nd the dt rte for the RFnet is 1 Mbps. For the ACnet, we ssume n coustic communiction rnge of 3 km, dt rte of 3 bps using DSSS/MFSK, nd signl propgtion speed of 15 m/s. For the TDMA slot configurtion, we consider tht 5 slots re llocted to the buoys nd s = 7 slots re llocted for AUVs. This result in TDMA frme period of 12 slots nd when the slot time is 3 s, we obtin TDMA frme of 36 s. The reson for choosing 3 s is s follows. Every AUV nd buoy periodiclly exchnge becons in its llocted slot. We consider becon size is 32 bytes. An AUV becon contins the AUV ID, flight time, speed, bering, loction coordintes nd one or more fields used for ACK nd dt exchnge. A Buoy becon contins buoy ID, GPS-bsed loction nd time informtion nd one or more fields used for dt exchnge. We choose the TDMA slot size of 3 s to include the becon trnsmission time in the ACnet of.85 s nd 3 km propgtion time of 2 s. Note tht these ttributes re vribles in our code nd their vlues cn be chnged ccording with the plnned scenrio. The scenrio considered is presented in Section 3. Once n AUV is deployed, it gets to the surfce nd sends n initiliztion messge to the ship over the RFnet. The ship responds with messge contining the coordintes of the surveillnce region nd the TDMA slot llocted to this AUV. The ship does the ssignment bsed on the current AUV loction. It will select the closest region tht hs TDMA slot vilble. The AUV then dives nd performs the surveillnce tsk. We ssume the AUV speed of 3 m/s. We used the DSR routing protocol [7] for RFnet communiction. At the trnsport lyer, we used TCP protocol. We chose TCP protocol becuse it offers connection-oriented nd relible communiction service. Also, the congestion control mechnism prevents congestion when multiple AUVs trnsmit long messges (imges) to the ship. We used the DSR nd TCP protocol models provided by Opnet. The communiction in the ACnet uses TDMA, s we described in the previous sections. The prmeters we vried in our simultion re the number of AUVs, the ship loction, the trget detection intervl nd the number of imge messges sent by n AUV to the ship for every trget detected. The performnce metrics we mesured re the AUV to ship messge dely, ship to AUV commnd dely, sttus-updte messge dely (with or without the AUV becon dely), nd ggregte throughput. In the first set of mesurements we vry the number of AUVs deployed, between 1 nd 12. We ssume the ship is positioned t the center of the surveillnce rectngle. Trget detection time is uniformly distributed between 3 min nd 1 h nd the intervl the ship will issue new commnd for every AUV is uniformly distributed between 2 nd 4 min. In Fig. 4, we mesure the AUV to ship messge dely. This is for the cse when n AUV detects trget nd gets to the surfce to trnsmit the first imge to the ship. This dely is therefore mesured over RFnet. The size of n imge messge is 25 kbytes. Our results indicte n verge dely of 1 s nd the t 2.5 s. These vlues re dependent on the number of hops between AUV nd

9 M. Crdei / Computer Communictions xxx (26) xxx xxx 9 3 b AUV to ship messge dely, sec AUV count Ship to AUV commnd dely, sec AUV count Fig. 4. Messge delys in communiction between AUV nd ship. ship nd re lso ffected by MAC lyer contention, TCP nd routing overhed nd retrnsmission delys. In Fig. 4b, we mesure the dely for sending commnd from the ship to the AUV. Here the AUV is submerged, so this dely reflects both the RF nd coustic dely. The dely hs n verge of 2 s, with the t 95 s. The RFnet dely is ffected by the fctors mentioned before. Regrding the ACnet communiction, the commnd is sent by the buoy during its slot, so the commnd will be queued t the buoy between nd 33 s. As we send only one commnd per becon, it my hppen tht commnd is delyed few TDMA frme periods, especilly for cses with more AUVs. In our model, buoys re responsible for periodiclly sending sttus-updte messges to the ship with informtion bout AUVs in their rnge. Buoys collect this informtion from the AUV becons received in the ACnet. The frequency of sending informtion updtes by the buoys cn be set up to different vlues. For these simultions we ssume sttus updtes re sent every TDMA frme intervl, tht is, every 36 s. In Fig. 5, we represent the dely of the sttus-updte messges, wheres in the Fig. 5b, the dely is mesured from the moment Aggregte messge throughput t the ship, bps AUV count Fig. 6. Aggregte messge dely t the ship depending on the number of AUVs. it ws sent by the AUV. This dely includes the time the AUV becon ws sent over the ACnet (3 s), the dely occurred in the buoy (up to 33 s) nd the dely over the RFnet. In Fig. 6, we present the ggregte throughput mesured t the ship. As expected, the verge vlue increses with the number of AUVs..14 b Sttus-updte messge dely, sec AUV count AUV becon plus sttusupdte messge dely, sec AUV count verge verge Fig. 5. Sttus-updte messge delys.

10 1 M. Crdei / Computer Communictions xxx (26) xxx xxx AUV to ship messge dely, sec Centrl positioning Lterl positioning b Ship to AUV reply dely, sec Centrl positioning Lterl positioning Ship positioning Ship positioning Fig. 7. Messge dely depending on the ship positioning. In the next set of mesurements, we study the impct of the ship loction on the messge delys between the AUV nd ship nd sttus-updte messge dely. We compred the cse when the ship is positioned t the middle of the surveillnce rectngle (e.g., close to B 6 in Fig. 3), versus the cse when it is positioned lterlly (e.g., close to B 1 in Fig. 3). A centrl ship positioning will reduce the verge number of hops involved in RFnet communiction Sttus-updte messge dely, sec Centrl positioning Ship positioning Lterl positioning Fig. 8. Becon dely depending on the ship positioning. between ship nd buoys or surfce AUVs. In Fig. 7, we mesure the dely of the messges with 25 KB imges, sent from AUVs to ship. Fig. 7b represents the reply sent from the ship to the surfce AUV. This is 96 bits pcket which is used for ACK, or for issuing new commnds to the AUV or for requesting for more pictures. Next, Fig. 8 illustrtes the vrition of the sttus-updte messge depending on the ship loction. In the next test, we study the impct of vrying the rte t which AUVs detect trgets: 15, 3, 45, nd 6 min. The dely of the messges with pictures sent from AUVs to the ship is represented in Fig. 9. Also, in Fig. 9b, we represent the ggregte throughput. As more messges re generted with smller trget detection intervl, the ggregte throughput increses correspondingly. In the lst set of experiments, we vry the number of messges sent by AUVs per trget detected. We ssume tht when trget is detected, fter the AUV sends the first imge messge, the ship lwys sks for ll the remining imges. The gol is to test the system comportment when vrying the lod. In this cse we set the trget detection intervl s being uniformly distributed between 2 nd 3 min. We represent in Fig. 1, the dely of the messges with pictures sent by the AUVs to the ship nd in Fig. 1b, the dely of the sttus-updte messges. We observe n AUV to ship messge dely, sec b Aggregte throughput t the ship, bps Trget detection intervl, min Trget detection intervl, min Fig. 9. Messge dely nd ggregte throughput bsed on the trget detection intervl.

11 M. Crdei / Computer Communictions xxx (26) xxx xxx 11 3 b AUV to ship messge dely, sec Number of messges sent by AUV per trget detected Sttus-updte messge dely, sec Number of messges sent by AUV per trget detected Fig. 1. AUV to ship nd sttus-updte messge dely depending on the number of messges sent by AUV per trget detected. Aggregte throughput t the ship, bps Number of messges sent by AUV per trget detected Fig. 11. Aggregte throughput t the ship depending on the number of messges sent by AUV per trget detected. increse in dely when the number of messges sent per trget detected is 1. In the lst grph, in Fig. 11, we represent the increse in the ggregte throughput with the number of messges sent by AUV per trget detected. 8. Conclusions In this pper, we ddress the problem of designing n energy-efficient scheduling mechnism nd sclble hybrid RF-coustic communiction rchitecture, tht fcilittes trnsfer of dt from AUVs to centrl control nd commnd sttion on ship. Communicting buoys serve s gtewys between the coustic nd RF networks nd provide messge forwrding cross the d hoc RFnet. Mjor impediments ssocited with underwter coustic networks tht were considered re severe bndwidth limittions nd long propgtion delys. The Medium Access Control scheduling method dopted is TDMA, providing collision-free, energy-efficient communiction on the low-dt rte coustic chnnel. We design sclble TDMA slot lloction mechnism with sptil-reuse chrcteristics. We lso design sclble buoy positioning mechnism tht gurntees every point in the surveillnce re is within communiction rnge of t lest three buoys for fcilitting y L U 1 U 2 U' 2 C B 2 B 4 B 6 B 8 O E b B B 1 3 B D 5 B 7 R W x L' 2 L 3 Fig. 12. L nd W computtion.

12 12 M. Crdei / Computer Communictions xxx (26) xxx xxx AUV underwter tringultion. Performnce of our communiction system is nlyzed with simultions in the Opnet network simultor. Acknowledgement The uthor thnks Dr. Andres Folleco for his vluble input on shllow wter coustic communictions. Appendix A. In this ppendix we compute the length L nd width W of the surveillnce rectngle, considering n buoys B 1,B 2,...,B n, n P 3, the p coordintes of B 2 s being B 2 (R/2,d y ), nd d y 2½; R ffiffiffi 3 =2Š. For the nottions we refer to Fig. 12 where n =8. From nu 2CL 2: cos ¼ W nd from nb 2U 2 E 3 DB 6 : cos ¼ 3R=2 2EB 6, therefore W ¼ 3R U 2 E 2 EB 6. In nb 3 DB 6 : EB 2 6 ¼ 1 4 ðd2 y þð3r 2 Þ2 Þ¼ 1 16 ð4d2 y þ 9R2 Þ nd in nu 2EB 6 : rffiffiffiffiffiffiffiffiffiffiffiffi U 2 E2 ¼ 1 16 ð7r2 4d 2 y Þ, therefore W ¼ 3R 7R 2 4d 2 y. 2 9R 2 þ4d 2 y Let us next compute L. L cn be written s L =2Æpr(B 3 U 1,Ox)+(n 5) * R/2, where pr(b 3 U 1,Ox) is the projection of B 3 U 1 on the Ox xis. We hve pr(b 3 U 1,Ox) =pr(b 3 U 2,Ox) =R Æ cos( + b), nd cosð þ bþ ¼cos cos b sin sin b ¼ 3R=2 2EB 6 EB 6 dy R 2EB 6 U 2 E R rffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffi ¼ 3 dy 7R 2 4d 2 y 7R. So we get L ¼ 3R=2 d 2 4d 2 y 4 2R 9R 2 þ4d 2 y y 9R 2 þ4d 2 y rffiffiffiffiffiffiffiffiffiffiffiffi 7R þðn 5ÞR=2 ¼ðn 2ÞR=2 d 2 4d 2 y y, for ny n P 3. References 9R 2 þ4d 2 y [1] Deprtment of Ocen Engineering t Florid Atlntic University, < [2] Center for Acoustic nd Vibrtions t FAU, < [3] H. Blkrishnn, V.N. Pdmnbhn, S. Seshn, R.H. Ktz, A comprison of mechnisms for improving TCP performnce over wireless links, IEEE/ACM Trnsctions on Networking (1999). [4] J. Ctipovic, D. Brdy, S. Etchemendy, Development of underwter coustic modems nd networks, Ocenogrphy 6 (1993). [5] X. Cheng, A. Theler, G. Xue, D. Chen, TPS: time-bsed positioning scheme for outdoor wireless sensor networks, IEEE INFOCOM (24). [6] J. Broch, D.A. Mltz, D.B. Johnson, Y. Hu, J. Jetchev, A performnce comprison of multi-hop wireless d hoc network routing protocols, in: ACM/IEEE Interntionl Conference on Mobile Computing nd Networking, October [7] The Dynmic Source Routing Protocol for Mobile Ad Hoc Networks (DSR), IETF Internet drft, Februry 22. [8] Hrdimnker, J.L.R. LeBlnc, P. Beujen, M. Singer, C. Boubli, G.T. Strutt, Chirp FSK modem for high relibility communiction in shllow wter, IEEE Ocens 99 (1999). [9] < [1] J. Rice, B. Creber, C. Fletcher, P. Bxley, K. Rogers, K. McDonld, D. Rees, M. Wolf, S. Merrim, R. Mehio, J. Prokis, K. Scussel, D. Port, D. Green, Evolution of Seweb underwter coustic networking, in: Proceedings of the IEEE Ocens Conference, Providence, RI, September 2. [11] F. Slv-Gru, M. Stojnovic, Multi-cluster protocol for d hoc mobile underwter coustic networks, in: Proceedings of the IEEE OCEANS 3 Conference, September 23. [12] B. Shrif, J. Neshm, O.R. Hinton, A.E. Adms, A computtionlly efficient Doppler compenstion system for underwter coustic networks, IEEE Journl of Ocenic Engineering 25 (1) (2). [13] E.M. Sozer, M. Stojnovic, J.G. Prokis, Underwter coustic networks, IEEE Journl of Ocen Engineering 25 (1) (2) [14] Wireless Outdoor Adpter, < usb_outdoor.pdf/>. [15] J. Prokis, E. Sozer, J. Rice, M. Stojnovic, Shllow wter coustic networks, IEEE Communictions Mgzine 39 (11) (21). [16] A. Tnenbum, Computer Networks, fourth ed., Prentice Hll, 23. [17] W. Stllings, Wireless Communictions nd Networks, Prentice Hll, 22. [18] G.G. Xie, J. Gibson, A networking protocol for underwter coustic networks, Technicl report TR-CS--2, Deprtment of Computer Science, Nvl Postgrdute School, December 2. Mihel Crdei is Assistnt Professor in the Deprtment of Computer Science nd Engineering t Florid Atlntic University, nd Director of the NSF-funded Wireless nd Sensor Network Lbortory. Dr. Crdei received her Ph.D. nd M.S. in Computer Science from the University of Minnesot, Twin Cities, in 23 nd 1999 respectively. Her reserch interests include wireless networking, wireless sensor networks, network protocol nd lgorithm design, nd resource mngement in computer networks. She is member of IEEE nd ACM.