Acoustic MIMO Communications in a Very Shallow Water Channel

Transcription

1 J. Marie Sci. Appl. (2015) 14: DOI: /s Acoustic MIMO Couicatios i a Very Shallow Water Chael Yuehai Zhou, Xiulig Cao ad Feg Tog * Key Laboratory of Uderwater Acoustic Couicatio ad Marie Iforatio Techology of the Miister of Educatio, Xiae Uiversity, Xiae , Chia Abstract: Uderwater acoustic chaels pose a great difficulty for the developet of high speed couicatio due to highly liited bad-width as well as hostile ultipath iterferece. Elighteed by rapid progress of ultiple-iput ultiple-output (MIMO) techologies i wireless couicatio scearios, MIMO systes offer a potetial solutio by eablig ultiple spatially parallel couicatio chaels to iprove couicatio perforace as well as capacity. For MIMO acoustic couicatios, deep sea chaels offer substatial spatial diversity aog ultiple chaels that ca be exploited to address siultaeous ultipath ad co-chael iterferece. At the sae tie, there are icreasig requireets for high speed uderwater couicatio i very shallow water area (for exaple, a depth less tha 10 ). I this paper, a space-tie ultichael adaptive receiver cosistig of ultiple decisio feedback equalizers (DFE) is adopted as the receiver for a very shallow water MIMO acoustic couicatio syste. The perforace of ultichael DFE receivers with relatively sall uber of receivig eleets are aalyzed ad copared with that of the ultichael tie reversal receiver to evaluate the ipact of liited spatial diversity o ulti-chael equalizatio ad tie reversal processig. The results of sea trials i a very shallow water chael are preseted to deostrate the feasibility of very shallow water MIMO acoustic couicatio. Keywords: uderwater acoustic; uderwater acoustic couicatio; ultiple-iput ultiple-output (MIMO); decisio feedback equalizer (DFE); very shallow water; ulti-chael; tie reversal Article ID: (2015) Itroductio 1 There is rapidly icreasig R&D iterest i high data rate uderwater acoustic couicatio systes i the fields of oceaic exploitatio, uderwater costructio, oceaography research, ad atioal defese (Chitre et al., 2008). Features of uderwater acoustic chaels (Li ad Preisig, 2007; Rouseff et al., 2009), such as arrow badwidth, serious ultipath, Doppler spread, ad Received date: Accepted date: Foudatio ite: Supported by the Natioal Natural Sciece Foudatio of Chia (Nos , ) ad the Ope Project Progra of the Key Laboratory of Uderwater Acoustic Sigal Processig, the Miister of Educatio (Southeast Uiversity) (No. UASP1305). *Correspodig author Eail: ftog@xu.edu.c Harbi Egieerig Uiversity ad Spriger-Verlag Berli Heidelberg 2015 backgroud oise are recogized as key liitatios for R&D of high data rate uderwater acoustic couicatios (Sog et al., 2008; Cotter ad Rao, 2002; Wu et al., 2013; Stojaovic, 2008, Zeg et al., 2010). Elighteed by sigificat success of ultiple-iput ultiple-output (MIMO) techology i wireless couicatio fields, sigificat data rate icreases ca be achieved uder acoustic chaels by siultaeously trasittig ultiple data streas fro a bak of trasitters. Previous ivestigatio ad experiets have idicated that, MIMO systes are capable of usig ultiple spatially parallel uderwater couicatio chaels to iprove couicatio perforace icludig capacity (Sog et al., 2007; Lig et al., 2014; Tao et al., 2010). Multichael decisio feedback equalizer (DFE) receivers ad tie reversal receivers are widely ivestigated i the research couity as coheret receivers for MIMO acoustic couicatio. A space-tie equalizer cosistig of ultiple DFE equalizers is adopted as the coheret receiver for MIMO acoustic couicatios (Fly et al., 2004; Sog ad Ritcey, 1996). I (Sog et al., 2011; Yag, 2005), a low coplexity tie reversal receiver is proposed by cobiatio of ultiple tie reversal processors ad sigle chael equalizers. I (Zhou et al., 2014), a tie reversal receiver ad a space-tie receiver are joitly adopted as a selective tie reversal receiver to facilitate low coplexity ipleetatio ad selective focusig of chael with a log delay spread. However, ost ivestigatios o MIMO acoustic couicatio are carried out i uderwater acoustic chaels with a large depth (>100 ) (Sog et al., 2006), which offers substatial spatial diversity for exploitatio as well as eablig the deployet of large receivig arrays (uber of eleet>10). Give icreasig high speed couicatio requireets i very shallow water such harbors, bridges, ad coastal facilities, the feasibility ad perforace of MIMO techology i very shallow chaels, i.e., with a depth of saller tha 10, is worth further aalysis ad validatio. This paper presets the ipleetatio ad perforace evaluatio of a ultichael DFE receiver for very shallow acoustic MIMO couicatio, so as to accoodate the liited spatial diversity i a very shallow water chael.

2 2 Yuehai Zhou, et al. Acoustic MIMO Couicatios i a Very Shallow Water Chael Cosiderig the extree shallow water depth, that excludes the deployet of a large receivig vertical array, a sall size receiver array, with oly 2, 3 ad 4 eleets, is aalyzed ad copared to evaluate the ipact of spatial diversity o the perforace of ulti-chael equalizer ad the utichael tie reversal receivers. At the sae tie, the capabilities of two types of receiver, to suppress ultipath iterferece ad co-chael iterferece of MIMO chael, with sall uber of receivers ad sall size DFE, are ivestigated. 2 MIMO odel ad receiver structure 2.1 Syste odel of MIMO acoustic couicatio The classic odel of a MIMO acoustic couicatio syste with N trasitters ad M receivers ca be writte as (Sog et al., 2011; Lig et al., 2014): N L1, 1 l0 y k s k lh k,lz k (1) where, y k ad z k are the receivig sigal ad additive oise at the th receiver, respectively. s (k) ad h, (k,l) are the trasittig sigal of the th trasitter ad the chael ipulse respose betwee -th couple, respectively. k is tie idex for observatio tie, l is tie idex for tie delay, L is the tie delay diesio of the chael ipulse respose. Uder the assuptio the chael reais stable i P saples, Eq. (1) ca be expressed as: N y A h z (2) 1 where the P row-l colu atrix, A is: sk L sk L1 sk 1 sk L1 sk L sk 2 A sk LP1 sk LP2 sk P (3) with: y 1 1 T y k L y k L y k LP s 1 1 T s k L s k L s k h T, h, k, h, k, h, k,l z 1 1 T z k L z k L... z k L P (4) Eq. (2) ca be further expressed as: where, y Ah w (5) A A, A, A h h h h T 1 2 N 1, 2, N, The MIMO chael h ca be estiated with least square (LS) or iiize ea square error (MMSE) ethod. For a ulti-chael DFE receiver, the effects of uderwater acoustic chaels are addressed i the for of teporal-spatial equalizatio, which is updated by adaptive algoriths such as RLS ad LMS to track the tie variatios of the acoustic chaels. As the tie reversal receiver geerally eeds a large uber of receivig eleets (>10) to achieve eaigful perforace (Yag, 2005), the ulti-chael DFE receiver is suitable for very shallow chaels due to its tolerace of sall size array with sall uber of eleets. 2.2 The structure of ultichael DFE receiver The classic ultichael DFE receiver for MIMO couicatio is illustrated i Fig.1 (Fly et al., 2004; Zhou et al., 2014). Fig. 1 Illustratio of the ultichael DFE receiver As show i Fig. 1, the ulti-chael DFE cosists of N forward filters (FF) desiged to recover N trasittig sequeces, each of which is coposed of MK f -order FIR filters associated with M receivers. I Fig1, W k w1 k w2k wn k deotes M(K f +1) N coefficiet atrix of N forward filters, where w (k) correspods to the th (1 N) FF to recover the th trasittig sequece, the carrier phase of which is j copesated by the e ter drive with a secod order phase lock loop (PLL). p k is the output of the th FF, expressed as: H j p k w ky ke (6) where y k is M(K f +1) order iput vector of the th FF, expressed as: where, k k k kk T y y y 1 y f (7) yk y k y k y k 1 2 M The feedback filter (FB) cosists of a NK b N diesio

3 Joural of Marie Sciece ad Applicatio (2015) 14: 3 vector Bk b1k b2k bn k, where K b is the order of the FB filter, b (k)(1 N) is a NK b diesio vector. Thus, output of the th FB filter is: q kb H ks k d (8) ˆ where ˆs k dˆs ˆ ˆ T 1 k d s2 k d sn k d is NK b diesio iput vector of the FB filter, i.e., previously detected bits, d is a costat for delay copesatio. Thus, the th (1 N) layer variable vector z k produced by the FF ad FB filter is writte as: z k=p k+qk (9) The adaptive algoriths, such as RLS ad LMS, ca be used to update the DFE to accoodate tie variatios iduced by chaels. However, eough spatial diversity betwee ultiple chaels of the receivig array is recogized as a precoditio for satisfactory perforace of ulti-chael equalizatio. This is easily realized for uderwater acoustic chaels with eough depth but ay be difficult for very shallow chaels. Thus, oe of the key challeges for ulti-chael DFE MIMO receivers i very shallow water is to achieve equalizatio i the presece of liited spatial diversity. 2.3 The structure of ultichael tie reversal receiver For the classic tie reversal receiver, the chael resposes obtaied with the chael estiatio algorith are used to costruct a ultichael tie reversal receiver (Sog et al., 2011; Yag, 2005) expressed as: ˆs ˆ, l ˆ,l ˆ, y h s h z h,l s ˆ h,lhˆ,lz ˆ h,l s z hˆ l, (10) where h,l is the chael respose obtaied with chael estiatio algoriths, such as LS algorith or atchig pursuit algorith (MP) (Sog et al., 2011). By suig output of ulti-chael tie reversal processors, spatial diversity ca be achieved i the for of a ulti-chael tie reversal receiver as: M ˆ, 1 ˆs s (11) By couplig ultichael tie reversal with a sigle-chael adaptive DFE, the tie reversal receiver is capable of iprovig the adaptability to tie varyig chaels (Sog et al., 2011; Yag, 2005). The purpose of the sigle-chael DFE is to address the residual ISI ad accoodate the teporal variatio of the physical chael. The priciple of the sigle DFE atches a DFE filter i the ultichael-dfe receiver. However, it is oted that the tie reversal processor geerally requires a vertical array with a large uber (>10) to yield satisfactory spatial diversity (Yag, 2005). 3 Experiet i a very shallow water chael I this sectio, at-sea experiet results, i a very shallow water chael, are preseted to evaluate the perforace of the ultichael DFE MIMO receiver ad the ultichael tie reversal receiver. I the experiet cofiguratio, with the sae two trasittig sources, differet uber of receivig eleets, i.e., 2TX-4RX, 2TX-3RX ad 2TX-2RX, MIMO acoustic couicatio systes are adopted for the perforace evaluatio ad copariso with respect to spatial diversity. The odulatio forat was quadrature phase-shift keyig (QPSK) with a bit rate of 8 kilobits per secod ad a carrier frequecy of 16 khz. The badwidth of the trasducer couplig was khz. Origial saplig rate of the received data is 96 ksps. Saplig iterval of the basebad sequece is 1/2 of the sybol duratio. The MIMO acoustic couicatio experiet was carried out i a very shallow water acoustic chael at Wuyua bay, Xiae, Chia. The depth of the experiet area was about 6 at the tie of our MIMO couicatio experiet. The trasittig couplig was suspeded to depth of 2 ad 4 fro a boat, with the 4-eleet receivig vertical array suspeded to a depth rage of with a spacig of 1.25 at the pier (as show i Fig. 2(a)), to produce ulti-chael sigals for 4-chael, 3-chael ad 2-chael MIMO sigal processig. The uber of each MIMO trasitter (TX1, TX2) ad each eleet of the vertical receivig array (RX1, RX2, RX3, RX4) are also arked i Fig. 2(a). The soud velocity gradiet of the experiet chael is provided i Fig. 2(b). As the depth of the water is very sall, variatio of soud velocity alog the vertical array is tiy. The distace betwee the trasitter ad receiver is 1 000, correspodig to a SNR of 12 db for receivig sigals. (a) Experietal cofiguratio (b) Soud velocity gradiet Fig. 2 Experietal cofiguratio of MIMO acoustic couicatio syste

4 4 Yuehai Zhou, et al. Acoustic MIMO Couicatios i a Very Shallow Water Chael The MIMO chael ultipath respose, with respect to tie obtaied durig the experiet, is show i Fig. 3, fro which oe ca see that the very shallow water chaels cotai various ultipath copoets. As the depth is very shallow, the respose of all the MIMO chaels geerally exhibit a siilar ultipath patter, correspodig to the very liited spatial diversity that ca be exploited by the ulti-chael equalizer. (a) TX1-RX1 (c) TX1-RX3 (e) TX2-RX1 (b) TX1-RX2 (d) TX1-RX4 (f) TX2-RX2 (g) TX2-RX3 (h) TX2-RX4 Fig. 3 Chael respose with respect to tie I the MIMO couicatio sigal processig, the ultichael DFE receiver adthe tie reversal receiver, adopt 2 chaels, 3 chaels ad 4 chaels for ulti-chael processig respectively. For the ultichael DFE receiver, the adaptive DFE is updated with RLS algorith. I the sigal frae, the legth of the traiig sequece is 500 to fiish the traiig of RLS algorith, after which the DFE receiver is adaptively updated with the decided output. For the purpose of couicatio perforace evaluatio, five packets, each of which cotais bits, are used for calculatig the bit error rate (BER). The filter legth of the RLS updatig forward ad backward is set as 24, 12, respectively, with a RLS forgettig factor of The carrier phase is tracked with a secod-order PLL (phase lock loop), ebedded i the DFE, which is For the tie reversal receiver, the legth of the tie reversal processor is the sae with that of the chael estiator, set as 60. The atchig pursuit algorith (Sog et al., 2011) is adopted for perforig chael estiatio. The sigle chael adaptive DFE followig the ultichael tie reversal processor is updated with RLS algorith. The filter legth of the RLS updatig forward ad backward is set as 24, 12 respectively, with the RLS forgettig factor of The legth of the RLS traiig sequece is 500. The carrier phase is tracked with a secod-order PLL (phase lock loop) ebedded i the DFE, the PLL factor is The costellatio outputs correspodig to the MIMO ultichael DFE receiver as well as the tie reversal receiver with differet ubers of receivig eleets are provided i Figs. 4 ad 5 respectively, fro which oe ay see that, while the 2-chael DFE receiver is capable of yieldig preliiary equalizatio effects, the ultichael DFE receiver associated with a large uber of eleets achieves better separatio. I copariso, with the sae uber of receivig eleets, the tie reversal receiver achieves worse perforace copared to the ultichael DFE receiver does. The BER results obtaied by the two types of receivers with differet chaels are provided i Table 1. It idicates that, for BER of both TX1 ad TX2 data, icreasig the uber of receivig eleets cotributes to iprovig the BER perforace of both receivers, further validatig the role of spatial diversity i ulti-chael equalizer ad tie reversal. While the ulti-chael DFE receiver with 2 chaels achieves the BER, i.e., 0.06 for TX1 ad for TX2, the 2-chael tie reversal receiver correspods to the BER of 0.22 for TX1 ad for TX2, which is cosistet with the result of the costellatio plot. I additio, the output SNRs of the two types of receiver with respect to the uber of eleets are preseted i Table 2. As revealed by Table 2, both types of receivers with a large uber of eleets yield a higher output SNR tha receivers with a sall uber of eleets do. For the TX 2 case, the 2-chael, 3-chael as well as 4-chael DFE receiver ad tie reversal receiver produces a output SNR of 13.7 db, 16.0 db, 16.4 db, ad 12.2 db, 12.5 db, 13.7 db respectively. Meawhile, fro the MIMO couicatio perforace, as idicated by Figs. 4 5 ad Tables 1 2, it is evidet that the quality of MIMO chaels fro the 2d trasitter is superior to that fro the 1st trasitter.

5 Joural of Marie Sciece ad Applicatio (2015) 14: 5 The reaso why DFE is better tha TR i our experiet is that the tie reversal receiver geerally requires a large uber of vertical eleets (>10) to achieve spatial diversity (Yag, 2005). Ufortuately, the very shallow uderwater acoustic chael excludes the deployet of a vertical array with large uber of eleets, thus liitig the perforace of the TR receiver. (a) TX1, M=2 (b) TX1, M=3 (c) TX1, M=4 (d) TX2, M=2 Table 1 The BER perforace correspodig to differet uber of receivig eleets M=2 (RX1,3) M=3 (RX1,2,3) M=4 (RX1,2,3,4) MIMO Receiver TX1 TX2 ultichael DFE ultichael TR ultichael DFE ultichael TR ultichael DFE ultichael TR Table 2 The output SNR of DFE receiver correspodig to differet uber of receivig eleets db (e) TX2, M=3 (f) TX2, M=4 Fig. 4 Scatter plots of MIMO ultichael DFE receivers with differet uber of receivig eleets MIMO Receiver TX1 TX2 M=2 (RX1,3) ultichael DFE ultichael TR M=3 (RX1,2,3) ultichael DFE ultichael TR M=4 (RX1,2,3,4) ultichael DFE ultichael TR Coclusios (a) TX1, M=2 (b) TX1, M=3 (c) TX1, M=4 (d) TX2, M=2 I view of the requireet for high speed acoustic couicatio i very shallow water chaels, a ultichael DFE MIMO receiver ad a ultichael tie reversal receiver are ipleeted ad evaluated to verify the feasibility ad perforace of MIMO acoustic couicatio i very shallow water. The experiet results obtaied i a real very shallow water chael are preseted to show that the ultichael DFE receiver is capable of achievig MIMO couicatio with a relatively sall uber of chaels (for exaple, 4 is eough for achievig satisfactory equalizatio effect i our ivestigatio) give the presece of liited spatial diversity associated with very shallow water. Refereces (e) TX2, M=3 (f) TX2, M=4 Fig. 5 Scatter plots of MIMO ultichael TR receivers with differet uber of receivig eleets Chitre MS, Shahabodee S, Stojaovic M (2008). Uderwater acoustic couicatios ad etworkig: Recet advaces ad future challeges. Marie Techology Society Joural, 42(1): DOI: / Cotter SF, Rao BD (2002). Sparse chael estiatio via atchig pursuit with applicatio to equalizatio. IEEE Trasactios o Couicatio, 50(3), DOI: /

6 6 Yuehai Zhou, et al. Acoustic MIMO Couicatios i a Very Shallow Water Chael Fly JA, Ritcey JA, Rouseff D, Fox WLJ (2004). Multichael equalizatio by decisio-directed passive phase cojugatio: Experietal results. IEEE Joural of Oceaic Egieerig, 29(3), DOI: /JOE Li Weichag, Preisig JC (2007). Estiatio of rapidly tie-varyig sparse chaels. IEEE Joural of Oceaic Egieerig, 32(4), DOI: /JOE Lig Ju, Ta Xig, Yardibi T, Jia Li, Nordevaad ML, He Hao, Zhao Kexi (2014). O Bayesia chael estiatio ad FFT-based sybol detectio i MIMO uderwater acoustic couicatios. IEEE Joural of Oceaic Egieerig, 39(1), DOI: /JOE Rouseff D, Badiey M, Sog A (2009). Effect of reflected ad refracted sigals o coheret uderwater acoustic couicatio: Results fro the Kauai experiet (KauaiEx 2003). Joural of the Acoustical Society of Aerica, 126(5), Sog A, Badiey M, McDoald VK, Yag TC (2011). Tie reversal receivers for high data rate acoustic ultiple-iputultiple-ouput couicatio. IEEE Joural of Oceaic Egieerig, 36(4), DOI: /JOE Sog A, Badiey M, Sog HC, Hodgkiss WS, Porter MB, the KauaiEx Group (2008). Ipact of ocea variability o coheret uderwater acoustic couicatios durig the Kauai experiet (KauaiEx). Joural of the Acoustical Society of Aerica, 123(2), Sog BG, Ritcey JA (1996). Spatial diversity equalizatio for MIMO ocea acoustic couicatio chaels. IEEE Joural of Oceaic Egieerig, 21(4), DOI: / Sog HC, Hodgkiss WS, Kupera WA (2007). MIMO tie reversal couicatios. Proceedigs of the Secod Workshop o Uderwater Networks (WuWNet 07), New York, DOI: / Sog HC, Roux P, Hodgkiss WS, Kupera WA, Akal T, Steveso M (2006). Mutiple-iput-ultiple-output coheret tie reversal couicatios i a shallow-water acoustic chael. IEEE joural of Oceaic Egieerig, 31(1), DOI: /JOE Stojaovic M (2008). Efficiet processig of acoustic sigals for high-rate iforatio trasissio over sparse uderwater chaels. Physical Couicatios, 1(2), DOI: /j.phyco Tao Ju, Zheg YR, Xiao Chegsha, Yag TC (2010). Robust MIMO uderwater acoustic couicatio usig turbo block decisio-feedback equalizatio. IEEE Joural of Oceaic Egieerig, 35(4), DOI: /JOE Wu FY, Zhou YH, Tog F, Kaster R (2013). Siplified p-or-like costrait LMS algorith for efficiet estiatio of uderwater acoustic chaels. Joural of Marie Sciece ad Applicatio, 12(2), DOI: /s Yag TC (2005). Correlatio-based decisio-feedback equalizer for uderwater acoustic couicatios. IEEE Joural of Oceaic Egieerig, 30(4), DOI: /JOE Zeg Weju, Jiag Xue, Li Xili, Zhag Xiada (2010). Decovolutio of sparse uderwater acoustic ultipath chael with a large tie-delay spread. Joural of the Acoustical Society of Aerica, 172(2), Zhou Yuehai, Zeg Ku, Tog Feg, Che Youga (2014). Selective tie reversal receiver for uderwater acoustic MIMO couicatios. Proceedigs of MTS/IEEE OCEANS 2014, Taipei, Chia, 1-6. DOI: /OCEANS-TAIPEI