Abstract. 1. Introduction

Transcription

1 Wreless Sensor Network, 00,, do:0.436/wsn Publshed Onlne May 00 ( A New Method to Improve Performance of Cooperatve Underwater Acoustc Wreless Sensor Networks va Frequency Controlled Transmsson Based on Length of Data Lnks Abstract Vahd Tabataba Vakly, Mohammadjavad Jannat Iran Unversty of scence and technology, Tehran, Iran E-mal: vakly@ust.ac.r, mjannat@ee.ust.ac.r Receved March 5, 00; revsed Aprl 0, 00; accepted Aprl 9, 00 In ths paper a new method to mprove performance of cooperatve underwater acoustc (UWA) sensor networks wll be ntroduced. The method s based on controllng and mzng carrer frequences whch are used n data lnks between network nods. In UWA channels Pathloss and nose power spectrum densty (psd) are related to carrer frequency. Therefore, unlke rado communcatons, n UWA Communcatons sgnal to nose rato (SNR) s related to frequency besdes propagaton lnk length. In such channels an mum frequency n whole frequency band and lnk lengths cannot be found. In Cooperatve transmsson, transmtter sends one copy of transmtted data packets to relay node. Then relay dependng on cooperaton scheme, amplfes or decodes each data packet and retransmt t to destnaton. Recever uses and combnes both receved sgnals to estmate transmtted data. Ths paper wants to propose a new method to decrease network power consumptons by controllng and sub-mzng transmsson frequency based on lnk length. For ths purpose, underwater channel parameters s smulated and analyzed n km to 0 km lengths (mdrange channel). Then lnk lengths sub categorzed and n each category, mum frequency s computed. Wth these sub mum frequences, sensors and base staton can adaptvely control ther carrer frequences based on lnk length and decrease network s power consumptons. Fnally Dfferent Cooperatve transmsson schemes Decode and Forward (DF) and Amplfy and Forward (AF), are smulated n UWA wreless Sensor network wth and wthout the new method. In recever maxmum rato combner (MRC) s used to combnng receved sgnals and makng data estmatons. Smulatons show that the new method, called AFC cooperatve UWA communcaton, can mprove performance of underwater acoustc wreless sensor networks up to 40.4%. Keywords: Underwater Acoustc Communcatons, Wreless Sensor Networks, Cooperatve Transmsson, Decode and Forward, Amplfy and Forward. Introducton Recently, underwater acoustc wreless sensor networks (UWA-WLSN) become a hot topc n acoustc communcatons zone. Major dfference between ths knd of sensor networks and tradtonal ones s ther specal physcal layer whch effects on acoustc waves used to transmt data. Usng acoustc waves s not only but the best manner to acheve suffcent range and data rate n underwater envronment. The problem s that rado waves wll be absorbed soon n water and cannot support suffcent rang and data rate. Moreover, lght experences hgh dsperson n underwater envronment and agan cannot support suffcent range and rate. Unlke them, new progresses n under water acoustc communcatons make relable data lnks for several klometers concevable. So that researchers are effectvely encouraged gong ahead n underwater acoustc communcaton. Ths strange knd of physcal layer has several nfluences on channel parameters. Frstly, acoustc waves move slowly, about 500 m/s, n water whch s one ffth of rado waves speed n atmosphere []. So that acoustc Copyrght 00 ScRes.

2 38 V. T. VAKILY ET AL. waves have large delay spreads. In rado channels Pathloss only depends on lnk length. But acoustc waves experence frequency and lnk length dependent Pathlosses n underwater envronment. Therefore, Lnk s carrer frequency effects on ts total performance. Because of suspended partcles and small bubbles, acoustc waves are dspersed wdely n underwater envronment. Furthermore, reflectons from surface and bottom of sea ncrease channel fadng. All ponts mentoned before most be consdered n desgn of underwater acoustc wreless systems. Themes mentoned show that, lke rado communcaton, n UWA communcaton range and bandwdth are mportant bottlenecks. Observed nose n the ocean s categorzed nto two groups, man-made nose and ambent nose []. In deep ocean man made nose s gnorable, whereas, n presence of shppng actvtes or besdes shore man made nose ncreases level of total nose power. On the other hand, geysers, earthquakes, heat and some knds of marne anmals can be consdered as major sources of ambent nose. Total nose n underwater acoustc envronment s related to sgnal carrer frequency. In part 3., there are further descrptons and statstcal model of underwater acoustc nose. Snce pathloss and nose power are frequency dependent, SNR n underwater acoustc communcatons s related to frequency. Therewth, lke all wreless channels, n UWA channels, SNR s a functon of lnk length. Therefore SNR s nfluenced from two major parameters, lnk length and frequency. It means that, changes n length can nfluences on mum frequency of system. In Secton 4, a new method to ncrease system performance s descrbed. In ths method lnk lengths are subcategorzed. Then, for each category, mum frequency s defned. Fnally, usng proposed adaptve algorthm n chapter 4, all network nodes adjust ther carrer frequences to mze total network s performance. In ths paper, mentoned algorthm s called adaptve controlled frequency (ACF) method. Smulatons of Secton 4 shows that, n compare wth tradtonal method, ACF can ncrease system performance up to 9.7% Accordng to consderable progresses n rado communcatons, researches try to mprove UWA systems by applyng new schemes whch are lent from rado communcatons. One of these methods s cooperatve communcaton whch s sutable to use n wreless sensor networks. In chapter 5 two schemes of cooperatve communcaton, DF and AF, s adjusted, appled and smulated n UWA-WLSN. Smulatons show that Compared wth no cooperaton method, AF and DF methods can mprove system performance up to 7 and percents, respectvely. (Authors of the paper publshed ther Frst works on UWA cooperatve WLSN n [3] whch are summarzed n chapter 5). Usng results of prevous chapters, n chapter 6 a new method to mprove UWA communcaton s proposed. In ths method, whch s called ACF cooperatve UWA communcaton, dependng on lnk length between nodes and by performng ACF algorthm, mum frequency for each path s defned. Then data packets are transmtted on all paths. Fnally usng cooperatve schemes and MRC, receved sgnals combned and data packets are estmated. The remnder of ths paper organzed as follows. In Secton a bref lterature revew s presented. Next secton assgned to descrpton of UWA channel. AFC algorthm s proposed n Secton 4. Cooperatve UWA wreless communcaton descrbed and smulated n next secton. In Secton 6 AFC Cooperatve scheme n UWA- WLSNs s proposed and smulated. Fnally, whole work s summarzed and concluded n last secton. Ths paper s Smulatons show that ACF cooperatve UWA communcaton scheme can mprove performance of UWA-WLSNs up to 40.4%.. Lterature Revew Leonardo Da Vnc was the frst who tres to use underwater acoustc nformaton to detect shps. Wth a long tube submerged under the sea, he lstened to sounds whch were propagated from shps and detected them. The frst operatonal underwater acoustc (UWA) communcaton system was an underwater telephone, developed n 945 n the Unted States, for communcaton wth submarnes. It used a sngle sde-band suppressed carrer modulaton n the 8- khz band, and could operate over several klometers [4]. The development of dgtal communcatons for undersea applcatons dates back to smple png-based use of sonars that operate n the audble band [5]. Frst works on UWA multpath channels to ncrease data rate was reported by Ross Wllams and Henry Battestn n 97 [6]. Through the 980s phase coherent communcaton was used almost exclusvely for deep-water vertcal lnks, but n the early 990s phase coherent communcaton n multpath channels began to attract attenton, as ncoherent methods were lmted to a bandwdth effcency of approxmately 0.5 bts per Hz [4]. Snce the publcaton of the specal ssue on ocean acoustc data Telemetry n the IEEE journal of oceanc engneerng n 99, fundamental advances have been made n ths feld [4]. Bandwdtheffcent phase-coherent communcatons, prevously not consdered feasble, were demonstrated to be a vable way of achevng hgh-speed data transmsson through many of the underwater channels, ncludng the severely tme-spread horzontal shallow water channels [7-9]. The new generaton of UWA communcaton systems, based on the prncples of phase-coherent detecton technques, s capable of achevng raw data throughputs that are an order of magntude hgher than those of the exstng systems [0] whch are based on noncoherent detecton me- Copyrght 00 ScRes.

3 V. T. VAKILY ET AL. 383 thods. These results open many new possbltes for applcaton of UWA communcatons. Notable among the emergng applcatons s the concept of an autonomous oceanographc samplng network (AOSN) []. Ths network wll provde exchange of data, such as control, telemetry and vdeo sgnals between many network nodes. The network nodes, both statonary and moble ones, located on underwater vehcles and robots, wll be equpped wth varous oceanographc nstruments, such as hydrophones, current meters, sesmometers, sonars and vdeo cameras. Major dffcultes are encountered due to the long propagaton tmes n the underwater channels [3]. Frst protocols for acoustc local area networks (ALAN) have been proposed n [,3]. Throughput the 990s a number of addtonal systems were developed and commercalzed usng both coherent and noncoherent modulaton [5]. At hgh frequences approprate for shallow water communcatons, ray theory provdes the framework for determnng the coarse multpath structure of the channel [5]. As such a model does not capture the tme-varyng nature of the channel, efforts have been made to augment ths model wth a tme-varyng surface [4]. Some researchers model the shallow water channel as a Raylegh fadng channel but others challenge that assumpton, especally when dscrete arrvals can clearly be seen n the channel response. There has been no consensus among researchers on the model applcable n shallow waters. Recently, a ray theory based multpath model where the ndvdual multpath arrvals are modeled as Raylegh stochastc processes has been shown to descrbe the medum range very shallow water channel accurately [5]. Studes of acoustc propagaton through ansotropc shallow water envronments n the presence of nternal waves [6] may form the bass of future physcs-based channel modelng research. An addtve Gaussan nose assumpton s used commonly n the development of most sgnal processng and communcaton technques. Although ths assumpton s vald n many envronments, some underwater channels exhbt hghly mpulsve nose. Sgnal detecton [7] and Vterb decodng [8] technques developed for mpulsve nose models such as the symmetrc stable nose have been shown to perform better n warm shallow waters domnated by snappng shrmp nose [5]. A good revew of underwater network protocols can be found n [3]. A store-and-forward protocol was proposed n [9] for shallow-water ALAN s, where they use a form of packet rado network (PRN) protocol [0] that matches the shallow-water acoustc channel characterstcs. In [], the authors presented a clustered topology assumng full-duplex modems. Further and more detaled nformaton about recent advances n UWA communcatons and networks can be found n [5], where authors made comprehensve study on recent theoretcal advances, technologes and producton systems. 3. UWA Channel To specfy a specal wreless channel lke UWA channel several parameters must be defned. In ths chapter mportant parameters of UWA channels lke nose psd, pathloss and SNR s studed. 3.. Nose n UWA Channels Sources of Ambent nose n UWA channels can be categorzed and modeled n 4 groups. Nose power spectrum densty (psd) n underwater channel s depended on frequency, f, and can be modeled as []: N f N f N f N f N f () t s w th where N f 40 0 s0.5 6log f 60log f 3, s 0 0 Nw f log0 f 40log0 f 0.4, Nth f 5 0log 0 f and Nt f 7 30log 0 f. where Nt f, Ns f, Nw f and th N f are noses caused by turbulence, shppng actvtes, wnd and heat, respectvely. s s shppng actvty factor, whose value ranges between 0 and for low and hgh actvty, respectvely. And w s wnd velocty (0-0 m/s). Fgure shows smulated nose psd n 3 arbtrary cases, s 0, w 0 ; s 0.5, w 5 and s, w0, n frequences less than 00 khz. All other cases slde between these graphs. Fgure. Nose psd versus frequency n 3 arbtrary selectons of s and w. Copyrght 00 ScRes.

4 384 V. T. VAKILY ET AL. 3.. Pathloss and SNR n UWA Channels Sgnals n UWA Channels experence frequency and lnk length dependent pathloss whch s more complcated than rado channels and can be modeled as 3 T 0 log0 r0 r () where r s lnk length and absorpton coeffcent,, s functon of frequency. f 0. f 44 f f 400 f (3) Fgure shows f n frequences less than 00 khz. Frst part of () s smlar to rado channels and stands for power consumptons of sgnals whch are transmttng from source to destnaton n wreless channels. Second part corresponds to mechancal absorptons of travelng wave s power n underwater envronment whch s caused by mechancal nature of acoustc waves and specfes UWA channels. For an arbtrary sgnal power, by substtutng (3) n (), receved power n destnaton can be computed. Therefore, wth ad of () we have: SNRd, f 0logPT T 0logN (4) where PT s sgnal power, N s total nose power n transmsson band and d s lnk length. In Fgure 3 relatve SNR for several lnk length between 5 and 00 km smulated and plotted. It s obvous that 3dB bandwdth has nverse rato wth lnk length. Fgure 3 s an endorsement for dependency of mum frequency to lnk lent. It means that, an mum frequency cannot be found for whole frequences and ranges. In next secton ths problem s studed. Fgure 3. Relatve SNR versus Frequency n 5-00 km. 4. AFC Algorthm As t mentoned before, based on lnk length between transmtter and recever nodes, mum frequency dffers. Dependng on range of transmsson, mum frequency can be computed from max SNR d, f max 0 log P 0 log d dd 0 f d N f log f, s the fre- In (5), th mum frequency, quency whch maxmzes SNR d, f Therefore ths range. Fgure 4 shows d. T (5) when d d. f mzes transmsson performance n f at dfferent values of lnk length, Fgure. Absorpton coeffcent f n frequences less than 00 khz. Fgure 4. f at dfferent values of lnk length, d. Copyrght 00 ScRes.

5 V. T. VAKILY ET AL AFC n UWA Channels Table. f for d 0km and 0.5 km steps. Consderng prevous sectons, t s obvous that, transmtters and recevers whch are usng varable lnk lengths to communcate wth each other, cannot fnd an mum carrer frequency to use n all stuatons. To make UWA-WLSNs useable n dfferent geographc zones, topology of networks must be changeable. It means that, lnk length between nodes experence changes, whch s confned by network dmensons. Usng constant carrer frequency n whole network wll results n performance reducton. In ths part a new adaptve algorthm to decrease power consumptons of UWA communcaton whch are made by ths problem s proposed. Frst step of adaptve frequency controlled (AFC) algorthm s defnng lnk length nterval, d. Second step s calculatng mum frequency n each nterval. In ths part d s assgned equal to 500m and workng range of system s assumed, -0km, medum range. Wth these defntons there are 8 ntervals. For each nterval mum frequency can be calculated from max SNR d, f max 0 log P 0 log d ddd ddd 0 f d N f log where d d,,,...,max d T (6) d. Fgure 3. Shows that 3db frequency band s a decreasng functon of lnk length. Therefore, when d d d correspondng 3dB bandwdth of d s smallest n ths nterval. And we have max SNR d, f max 0 log P 0 log d ddd dd 0 f d N f 3 0 0log T (7) From (6) f for each nterval can be calculated. In Table, Usng (7), f for d 0km and d 0.5km s computed. Wth values of Table AFC algorthm can be performed as flowchart of Fgure 5. In Fgure 5 flowchart of AFC algorthm for UWA systems s shown. As t s seen, before startng algorthm several ntal defntons or calculatons most be done and a table lke Table must be formed. Then AFC algorthm can be started. Based on lnk length between transmtter and recever nodes, d must be defned and wth ad of Table carrer frequency can be found. Now one packet of data can be sent. If lnk length does not change more than d other pockets of data could be d (km) f (khz) d (km) f (khz) Fgure 5. Flowchart of AFC algorthm for UWA systems. sent. By changng lnk length more than d, d and carrer frequency must be defned agan. The algorthm contnues and all data packets wll be sent to recever node. In Fgure 6 result of applyng AFC algorthm n an UWA channel s shown. Smulatons of ths part prove that AFC algorthm can ncrease performance of UWA systems. As t can be seen n Fgure 6, n 0 km bt error rate (BER) decreases about 9.7% (from to 0.947). It means that AFC algorthm n UWA channels can ncrease system performance up to 9.7%. (Note that maxmum bt error rate s 0.5 and vertcal axs n Fgure 6 s graphed logarthmc). Copyrght 00 ScRes.

6 386 V. T. VAKILY ET AL. Fgure 7. UWA cooperatve channel. 5.. DF and AF n UWA Cooperatve WLSNs Fgure 6. Results of applyng AFC algorthm n an UWA channel n compare wth tradtonal scheme. 5. Cooperatve Communcaton Schemes n UWA Networks Concept of spacal dversty attracts attenton of wreless communcatons researchers and results n contnuous and massve quests to make use of t n WLSNs. In a wreless channel several paths can exst between transmtter and recever. If some of these paths are ndependent and have suffcent performance, channel performance can ncrease by sendng copes of data n these paths and combnng them n recever. Snce paths are ndependent total error probablty decreases. Therefore channel and system performance ncreases. Multple nput multple output (MIMO) systems make use of specal dversty by usng several antennas n transmtter and recever. These antennas must be separated enough to make correspondng paths ndependent. But In many usages of WLSNs t s mpossble. Because network nodes maybe smaller than they could support such separated antennas. To solve ths problem dea of cooperatve communcaton s proposed. In cooperatve communcaton systems, transmtter sends one copy of transmtted data packets to relay node. Then relay dependng on cooperaton scheme, amplfes or decodes each data packet and retransmt t to destnaton. If relay has proper poston, relay path wll be ndependent from drect path. Recever uses and combnes both receved sgnals to estmate transmtted data [3]. In Fgure 7 smplfed model of one relay UWA cooperatve channel, whch wll be used n contnuaton of the paper, s shown. In next secton two frequently used cooperatve schemes, DF and AF, n WLSNs s defned and appled to an UWA cooperatve WLSNs. Basc dea of DF s that one copy of data whch s sent to recever must be sent to relay too. In relay, ths message s decoded, corrected, coded and retransmtted to destnaton. In destnaton data whch s receved from both paths are decoded, corrected and combned to estmate transmtted message. Wth ths approach nether bt rate s achevable []. P( ) 3 R sup max I X ; Y X, I X, X ; Y (8) df It means that relay decodes message perfectly and retransmts t to destnaton. In Gaussan channels, f transmtter and relay send ther data coherently, above rate wll be achevable. In Fgure 7 transmtter, relay and recever nodes are called, and 3, respectvely. In Gaussan relay channels, channel gan between nodes and j (, j,, 3, j) s called h j. Receved sgnals n relay and recever experence addtve whte Gaussan nose wth unt power. Moreover power constrants n transmtter and recever are, E X P and E X P, respectvely. By computng (7), for Gaussan channel we have [3]: 0 R max mn log( h P), log h3 P h3 P h3 h3 PP (9) s a real constant and shows correlaton between X and X Whch are transmtted data from transmtter and relay, respectvely. If transmtter ) and relay ) cannot transmt coherently, correlaton s unusable and 0. Therefore we have 0 P 3 3 R max mn log h, log h P h P (0) Copyrght 00 ScRes.

7 V. T. VAKILY ET AL. 387 In AF scheme, relay node amplfes and retransmts receved sgnals wthout decodng t. Relay receves y ( b) n tme b. Then by consderng power constrants multples y ) by ( b. R max mnlog h P, 0 log h3 P h3 P () A Gaussan relay channel whch s modeled as comng expressons s consdered. Y h X Z Y h X Z Y h X Z () where Y 3 and Y 3 are receved sgnals from relay and transmtter. If all nose powers are unt and power constrants n transmtter and recever are P, we have: h P P (3) And AF can acheve bt rate h3 h P R log P h3 (4) h P h3 P If system works n low SNR regme Whch means P 0, We wll have h3 P R log P h3 (5) ln It means that n such stuaton, relay channel cannot help mprovng system performance and s unusable. Because n AF scheme both nose and power are amplfed and n low SNR AF cannot help data estmaton n recever. In Fgure 8 performances of one relay cooperatve UWA AF and DF channels s smulated and are compared wth no cooperaton mode. In ths fgure horzontal axs s dstance between transmtter and recever and vertcal axs s bt error rate (BER) whch s representatve of system performance. Relay poston s same as Fgure 7 and Ô, the angle between relay path and drect path n recever s 5. Fgure 8 shows that UWA cooperatve schemes can mprove performance of UWA-WLSNs. Maxmum mprovement s at 7500m where BER decreases from n noncooperaton mode to n DF mode and n AF mode whch means 33.38% and 7%, respectvely. (Note that maxmum bt error rate s 0.5 and vertcal axs n Fgure 8 s graphed logarthmc). Costs of such mprovements are relay establshment and correspondng source usages. Fgure 8. Performance of one relay UWA channel wth and wthout cooperaton. As t can be seen n Fgure 8, f length of drect path decreases performance mprovement of cooperatve schemes decreases too. When drect path decreases less than 5 km, relay path wll be longer than t. Therefore relay path experences larger Pathloss and cooperatve channel tends to weak relay channel. As t mentoned n last part, f relay channel s weak whch means experences low SNR, relay channel wll not help mprovng system performance and wll be unusable. 6. AFC Cooperatve UWA Algorthm In Sectons 4 and 5 AFC algorthm and Cooperatve schemes n UWA system are descrbed separately. In ths secton, AFC cooperatve UWA algorthm whch s a combnaton of cted methods n Secton 4 and 5 s proposed. 6.. AFC Cooperatve UWA-WLSNs In AFC cooperatve UWA method transmssons between all nodes (transmtters or relays) obey AFC algorthm whch s shown n Fgure 5. It means that for drect path and each part of relay path, carrer frequency must be defned by AFC algorthm and these frequences may dffer n dfferent lnk lengths. Therefore f AFC algorthm s appled n all parts of all paths, system may use dfferent workng frequences smultaneously and such networks are called AFC cooperatve UWA- WLSNs. In Fgure 9 AFC cooperatve UWA methods are smulated and compared wth other cted methods. Dependng on cooperaton scheme, AF and DF, new methods are called AF-AFC-UWA and DF-AFC-UWA methods, respectvely. Copyrght 00 ScRes.

8 388 V. T. VAKILY ET AL. Fgure 9. AFC cooperatve UWA schemes compared wth AF, DF and tradtonal schemes n an UWA-WLSN. In smulatons of ths part, based on proposed model n Fgure 7, transmtter power s 90 watts, relay poston s 5km away from recever, angle between drect path and relay path n recever, Ô, s 5 and drect path dffers from km to 0 km. In recever maxmum rato combner (MRC) s used to combne receved sgnals and make data estmatons. In Fgure 9, lke Fgure 8, when length of drect path decreases, performance mprovements whch are made by cooperatve schemes decrease too. In such stuatons, n spte of cooperatve schemes, AFC algorthm plays ts role and mproves system performance. Smulatons of ths part show that maxmum mprovement s happened n 8km where BER decreases form 0.80 n tradtonal scheme to n DF-AFC- UWA scheme whch means 40.4%. (Note that maxmum bt error rate s 0.5 and vertcal axs n Fgure 9 s graphed logarthmc). 7. Conclusons and Summares In Ths paper a new method to mprove performance of cooperatve UWA-WLSNs s proposed and evaluated va smulatons. The method s based on controllng and mzng carrer frequences whch are used n data lnks between network nods. In UWA channels Pathloss and nose psd are related to carrer frequency. Therefore, unlke rado communcatons, n UWA Communcatons SNR s related to frequency besdes propagaton lnk length. In such channels an mum frequency n whole frequency band and lnk lengths cannot be found. To solve ths problem AFC algorthm s proposed. In Cooperatve transmsson, transmtter sends one copy of transmtted data packets to relay node. Then relay dependng on cooperaton scheme, amplfes or decodes each data packet and retransmt t to destnaton. Recever uses and combnes both receved sgnals to estmate transmtted data. In ths paper, recever uses MRC to combne receved sgnals and make data estmatons. To use specal dversty of cooperatve communcaton n UWA-WLSNs these methods are appled and smulated. In frst secton of the paper, UWA communcaton s ntroduced. In Secton a summarzed lterature revew of UWA communcatons s offered. In subsequent secton, UWA channel ntroduced and some of ts parameters formulated. AFC algorthm for UWA channels s proposed n Secton 4. Then n next secton Cooperatve communcaton schemes, AF and DF, s presented and appled to UWA-WLSNs. Fnally, n Secton 6 AFC cooperatve UWA algorthm n WLSNs s proposed and evaluated wthn computer smulatons. Ths algorthm s a combnaton of cooperatve transmsson and AFC algorthm whch s proposed n prevous sectons. Smulatons show that the new method, called AFC cooperatve UWA communcaton, can mprove performance of underwater acoustc wreless sensor networks up to 40.4%. 8. Acknowledgments Authors would lke to express ther sncere thanks to Iran Telecommuncaton Research Center (ITRC) for ts valuable supports. 9. References [] M. Stojanovc, Underwater Acoustc Communcaton Channels: Propagaton Models and Statstcal Characterzaton, IEEE Communcatons Magazne, Vol. 47, No., January 009, pp [] M. Stojanovc, Underwater Acoustc Networks: Channel Models and Network Codng Based Lower Bound to Transmsson Power for Multcast, IEEE Journal on Selected Areas n Communcatons, Vol. 6, No. 9, December 008, pp [3] V. T. Vakly and M. J. Jannat, Performance of Cooperatve Transmsson n Underwater Acoustc Sensor Networks, ICCIA Internatonal Conference on Control Instrumentaton and Automaton, Tehran, 6-7 May 00. [4] M. Stojanovc, Underwater Acoustc Communcatons, IEEE Electro Internatonal, Boston, 5 May 995. [5] M. Chtre, S. Shahabudeen, L. Fretag and M. Stojanovc, Recent Advances n Underwater Acoustc Communcatons & Networkng, IEEE Journal of Oceanc Engneerng, 008, pp. -0. [6] R. E. Wllams and H. F. Battestn, Coherent Recombnaton of Acoustc Multpath Sgnals Propagated n the Deep Ocean, The Journal of the Acoustcal Socety of Amerca, Vol. 50, No. 6, 97, pp Copyrght 00 ScRes.

9 V. T. VAKILY ET AL. 389 [7] M. Stojanovc, J. A. Catpovc and J. G. Proaks, Phase Coherent Dgtal Communcatons for Underwater Acoustc Channels, IEEE Journal of Oceanc Engneerng, Vol. 9, No., January 994, pp [8] M. Stojanovc, J. A. Catpovc and J. G. Proaks, Adaptve Multchannel Combnng and Equalzaton for Underwater Acoustc Communcatons, The Journal of the Acoustcal Socety of Amerca, Vol. 94, No. 3, September 993, pp [9] M. Stojanovc, J. A. Catpovc and J. G. Proaks, Reduced-Complexty Multchannel Processng of Underwater Acoustc Communcaton Sgnals, The Journal of the Acoustcal Socety of Amerca, Vol. 98, No., August 995, pp [0] J. Catpovc. Performance Lmtatons n Underwater Acoustc Telemetry, IEEE Journal of Oceanc Engneerng, Vol. 5, No. 3, July 990, pp [] T. Curtn, J. Bellngham, J. Catpovc and D. Webb, Autonomous Oceanographc Samplng Networks, Oceanography, Vol. 6, No. 3, 993, pp [] D. P. Brady and J. A. Catpovc, Adaptve Multuser Detecton for Underwater Acoustcal Channels, IEEE Journal of Oceanc Engneerng, Vol. 9, No., Aprl 994, pp [3] J. L. Talavage, T. E. Thel and D. Brady, An Effcent Store-and-Forward Protocol for a Shallow Water Acoustc Local Area Network, Proceedngs of the Oceans Engneerng for Today s Technology and Tomorrow s Preservaton (OCEANS 94), Brest, 3-6 September 994, pp. I883-I888. [4] C. Bjerrum-Nese, L. Bjorno, M. A. Pnto and B. A. Quellec, A Smulaton Tool for Hgh Data-Rate Acoustc Communcaton n a Shallow-Water, Tme Varyng Channel, IEEE Journal of Oceanc Engneerng, Vol., No., 996, pp [5] M. Chtre, A Hgh-Frequency Warm Shallow Water Acoustc Communcatons Channel Model and Measurements, The Journal of the Acoustcal Socety of Amerca, Vol., No. 5, 007, pp [6] M. Badey, B. G. Katsnelson, J. F. Lynch and S. Pereselkov, Frequency Dependence and Intensty Fluctuatons due to Shallow Water Internal Waves, The Journal of the Acoustcal Socety of Amerca, Vol., No., 007, pp [7] M. A. Chtre, J. R. Potter and S. H. Ong, Optmal and Near-Optmal Sgnal Detecton n Snappng Shrmp Domnated Ambent Nose, IEEE Journal of Oceanc Engneerng, Vol. 3, No., 006, pp [8] M. A. Chtre, J. R. Potter and S. H. Ong, Vterb Decodng of Convolutonal Codes n Symmetrc -stable Nose, IEEE Transactons on Communcatons, Vol. 55, No., 007, pp [9] J. Jubn and J. D. Tornow, The DARPA Packet Rado Network Protocols, Proceedngs of the IEEE, Vol. 75, No., 987, pp. -3. [0] G. G. Xe and J. H. Gbson, A Network Layer Protocol for UANs to Address Propagaton Delay Induced Performance Lmtatons, MTS/IEEE Oceans 0 Conference, Boston, Vol. 4, 5-8 November 00, pp [] M. Stojanovc, On the Relatonshp between Capacty and Dstance n an Underwater Acoustc Communcaton Channel, ACM SIGMOBILE Moble Computng and Communcatons Revew, Vol., No. 4, October 007, pp [] T. Cover and A. E. Gamal, Capacty Theorems for the Relay Channel, IEEE Transactons on Informaton Theory, Vol. 5, No. 5, 979, pp [3] A. Host-Madsen and J. Zhang, Capacty Bounds and Power Allocaton for the Wreless Relay Channel, IEEE Transactons on Informaton Theory, Vol. 5, No. 6, 005, pp Copyrght 00 ScRes.