MOBILE ROBOT-BASED VIRTUAL V-BLAST MIMO TRANSMISSION SCHEME IN DISTRIBUTED WIRELESS SENSOR NETWORKS

Transcription

1 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: OBIL ROBOT-BASD VIRTUAL V-BLAST IO TRANSISSION SCH IN DISTRIBUTD WIRLSS SNSOR NTWORKS 1 WNBAI CHN, WI LI, 3 XIAOPIN ZHANG 1 School of Automaton, Bejng Informaton Scence and Technology Unversty, Bejng 10019, Chna Chna electroncs engneerng desgn nsttute, Bejng , Chna 3 State Key Laboratory of Informaton Photoncs and Optcal Communcatons, BUPT, Bejng , Chna ABSTRACT Amng at the requrements of hgh-speed data transmsson and low energy consumpton n energy constraned Wreless Sensor Networks (WSNs), a moble robot-based vrtual V-BLAST cooperatve IO transmsson scheme and an effcent approxmate maxmum lkelhood (L) detecton algorthm are proposed n ths paper. oblty of robot can reduce the communcaton dstance between the sensor nodes, IO-based WSNs can use ts dversty gan to overcome fadng effects and can also use ts multplexng gan to ncrease the data transmsson rate. As for the detecton algorthm n moble robotbased vrtual V-BLAST transmsson scheme, combned wth tradtonal decodng OSIC algorthm, the vectors needed to be detected n approxmate L algorthm are reduced. oble robot can receve wth ts own mult-antenna or usng collaboratve nodes around, and the energy effcency analyss of the two cases are all dscussed. Smulaton results show that the proposed scheme effectvely reduces network energy consumpton, and the effcent approxmate L detecton algorthm further mproves energy-effcency. Keywords: Wreless Sensor Networks(WSNs), oble Robot(R), ultple-nput multple-output antenna technology(io), Vertcal-bell layered space tme Archtecture (V-BLAST), axmum Lkelhood detecton(ld) 1. INTRODUCTION In emergency of dsasters, such as earthquake and fre, t s very sgnfcant to grasp the frst-hand nformaton on the scene of the accdent to mprove rescue effcency, to mnmze casualtes and losses[1].havng the features of wde coverage, strong adaptablty and flexble layout, Wreless Sensor Networks (WSNs) can quckly buld the network of nformaton gatherng, montorng and communcatons n the dsaster-strcken area, and the robot can form a flexble, effcent and fast rescue system under the support of the sensor network. So, t puts forward hgher requrements for WSNs to mprove the transmsson of data, voce and mage[1-]. Integratng the key technology of next generaton moble communcaton systems - ultple-nput ultple-output (IO) technology[3] nto WSN can make use of the dversty gan to overcome fadng effects and make use of the multplexng gan to mprove the nformaton transfer rate, thus to greatly mprove energy-effcency of the network[3-4]. As IO-based WSNs has become a new hotspot and moble robot has the capabltes of beng flexble and autonomous, ntroducng the robot technology nto WSNs can easly change the topology of WSNs and mprove the dynamc performance of the network[4]. By the powerful moble performance of oble Robot (R) and powerful communcaton performance of IO technology the energy effcency of WSNs can be mproved to enhance the overall performance of WSNs[5]. Apparently, upgradng WSNs performance wll further enhance the sensory abltes of the moble robot. The organc ntegraton of WSNs, R and IO wll make a contrbuton to buldng a powerful autonomous system amdst communcaton, measurement, control and mplementaton.. V-BLAST SYST Vertcal-bell layered space tme Archtecture (V- BLAST) system s the frst advent of IO experment system. Its structure s smple, practcal, and ts spectral effcency s strkng. V-BLAST encodng s very smple and ts complexty les manly n the sgnal detecton. Relevant documents 677

2 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: ndcate that the V-BLAST detecton has become a varety of research topcs[6-8]. There are commonly-used detecton algorthms lke lnear recever algorthm, sortng nterference cancellaton algorthm, QR decomposton algorthm and S algorthm. x 1 x x nt h 1 h 1nT h 11 h n R h 1 h nt h h n R n T h n R 1 r 1 r 1 Fgure 1: Transmsson model of IO system.1 Pont-Pont IO Wreless Systems Although encrypton can provde multmeda content wth the desred securty durng transmsson, once a pece of dgtal content s decrypted, the dshonest customer can redstrbute t arbtrarly[, 3]. As shown n Fgure 1, we consder a narrowband IO system channel model wth transmt and N receve antennas ( N,also denoted as N system). Under the assumpton of deal tmng and symbol-synchronous recever samplng, the N dmenson receved sgnal vector r = ( r1, r,..., r N ) can be modeled as where + t t t 1 m T x t,x t,...x t = t r nr r = Hx n (1) x denotes the dmensonal transmt sgnal vector, and satsfes H { xt x t } = PI m,where the superscrpt H stands for the Hermtan transpose. The average transmsson power of each transmsson antenna s set equal to P. n t s N dmensonal complex addtve whte Gaussan nose(awgn)vector, of whch each component s statstcally ndependent, and has zero mean and a varance σ / per dmenson. It satsfes { ntn t } = σ I. oreover, H s a IO H n n channel matrx wth the element of hj ( 1,1 j N ).ach element h j refers to the fadng between transmtter and recever j, whch s modeled as ndependent complex Gaussan random varables of equal varance and satsfes { j } 1 h =.. OSIC Decodng Algorthm of V-BALST The tradtonal V-BALST decodng algorthm OSIC can be descrbed brefly as follows: Intalzaton: =1 G1 = H + () Iteratve process: ( ) s = arg mn G (3) j { s1, s, Ks j 1} W = ( G ) (4) y xˆ S s = W r (5) T s S s ( s ) = Q y (6) = ˆ ( ) (7) r r x H s s G = H + (8) + 1 s = + 1 (9) Where, H + denotes oore-penrose pseudonverse of matrx H, ( G th ) s s the s column vector of G, H + s s pseudo-nverse of matrx H s whch th obtaned by elmnatng the s column of H. Functon Q ( ) denotes slcng operaton of the constellaton. quaton(3) denotes the operaton optmal orderng,whch s used to select the layer wth bggest post-detecton sgnal-to-nose rato to detect. quaton (5) denotes the operaton of nullng th all but the s transmtted symbol and (7) denotes the operaton of nterference cancellaton. V-BLAST frst decodes the strongest sgnal, then cancels the effect of ths strongest transmsson sgnal from each of the receved sgnals, and then proceeds to decode the strongest of the remanng transmsson sgnal, and so on. 3. OBIL ROBOT-BASD VIRTUAL V- BLAST IO TRANSISSION SCH 3.1 Vrtual IO Technque Cooperatve communcaton can be traced back to the research work about relay channel of Cover and I Gamal [9] n the 1970s. User cooperatve dversty concept was frst ntroduced by A. Sendonars et al n 1998 [10]. Almost at the same tme, Laneman et al proposed varous collaboratve strateges of fxed relayng n Ref.[4], selectve relayng based on channel measurement and enhanced relayng based on the feedback of termnal node. As a result, cooperatve communcaton has also become a hot research topc n cellular network, Ad-hoc network and other communcatons [6-9]. The basc dea of cooperatve communcaton s that usng one or more cooperatve partners of 678

3 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: moble termnals to forward copes of sgnals, whch consttutes a vrtual mult-antenna system. Compared wth mult-nput and mult-output system wth the tradtonal arspace, cooperatve dversty allows multple termnals wth a sngle antenna to share each other's antenna n a multuser envronment to form a vrtual multple antennas structure whch can obtan dversty gan. STBC-based cooperatve IO scheme mproves the energy effcency of the system by mprovng BR performance. Whle the scheme based on V-BLAST mproves the energy effcency by reducng the transmsson tme. In STBC coded cooperatve transmsson scheme mult-node at the transmtter sde makes full use of the mult-antenna transmsson dversty and obtans coded cooperaton dversty gan to overcome the channel fadng actually. On the other hand, multple sensor nodes smultaneously transmt ther data to the recevng sde n V- BLAST-based cooperatve transmsson scheme. Obvously, the scheme makes full use of the spatal multplexng gan and mproves transmsson rate. In general, STBC s more approprate to lowrate, hgh-relablty and low transmsson delay applcatons, whle V-BLAST-based scheme s more approprate to hgh-speed, low-relablty applcatons, such as audo/vdeo nformaton transmsson. V-BLAST-based cooperatve IO Transmsson was proposed by Jayaweera [1]. At the transmtter sde, The sensor nodes transmt ther own perceved data to the recever sde ndependently and smultaneously. Therefore, ths program s easer than STBC program mentoned above, because there are no jont encodng process at the transmtter sde. Launchng ther own perceved data ndependently and smultaneously to the recever sde means that the synchronzaton of data acquston nodes s a necessary condton that should be consdered n collaboratve IO transmsson whch s based on V-BLAST. Of course, the synchronzaton wll brng addtonal energy consumpton, but ths process can be completed by one leader node, whch s not constraned n the energy consumpton. Lke other References, ths paper assumes that data acquston nodes have been synchronzed, and ths part of the energy consumpton s gnored n the followng dscusson. Collectve Nodes Transmtter Collaboratve Nodes Recever... Snk Node Fgure : V-BLAST Collaboratve IO transmsson scheme The process of cooperatve recevng of the V- BLAST encodng s shown n Fgure. Snk node may requre multple secondary nodes around to be nvolved n recevng the sgnal to buld the recevng end of mult-antenna structure. Aded recevng node can send the receved sgnal to snk node n the next tme slot. Snk node makes an approprate choce of the decodng algorthm to obtan the sgnals transmtted by data acquston nodes, accordng to the sgnal receved from ts own antenna and the auxlary node. The results of Jayaweera S.K. show that the program can sgnfcantly mprove energy effcency n a multpath fadng envronment, and the more the number of collaboratve nodes has, the better energy effcency can be obtaned. 3. Dscussons on R-Based V-BLAST Transmsson Scheme oble robot can move at wll, and ts energy can be unrestrcted compared wth the common nodes of the sensor networks. So, usng oble robot as a relay or a Snk node, can beneft data acquston and prolong the lfetme of WSNs. R-Based V-BLAST transmsson scheme s shown n fgure 3. Accordng to the space constrant n applcaton, the robot can be mounted mult-antenna drectly or use multple secondary nodes around to buld vrtual mult-antenna. Fgure 3: R-Based V-BLAST Transmsson Scheme As shown n Fgure 3, mult-hop relay vrtual IO transmsson structure may be necessary n practcal applcaton of WSN, because of the coverage of sensor nodes. The nodes need to decode and to relay forward n ths transmsson scheme. In ths case, low complexty of the 679

4 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: decodng algorthm s nevtable, because of the poor calculatng ablty. In some applcatons of WSN, the physcal dmensons of the snk node can be unrestrcted, so t s possble to nstall multple antennas drectly, whch can further elmnate the need of local communcaton n recevng sde. Generally, the energy consumpton of snk node s not constraned, so the hgh BR performance of the recever algorthms has great sgnfcance n practce R servng as a snk node Generally, all the nodes are deployed statcly n WSN. The nodes close to the snk node are easy to become the bottleneck of the network, because they transmt forwards all data from other nodes and the power consumpton s very great. Accordng to the acquston task and geographc regons, the sensors can be grouped. Beng a moble node, the moble robot can determne the routes, the data acquston place and the tme to complete data acquston, accordng to the msson of data acquston and the maxmzaton of network s lfetme. Of course, the moble robot also can do data acquston when they move contnuously on a optmal moble path based on geographc traffc model. The communcaton dstance of the sensor nodes and the number of relay tasks s reduced n the scheme because of the movement of the moble robot. It also makes the bottleneck nodes be dstrbuted n the network evenly, whch can prolong the network s lfetme sgnfcantly. The problem to be consdered n the scheme s that the data store and data processng ablty of the moble robot. On some other real-tme occasons, such as fre detecton, mne rescue, the moblty of convergent node and the connectvty to the backbone network are all necessary. oreover, n order to transfer the percepton nformaton, the locaton of the moble robot s also necessary to all the sensor nodes. 3.. R servng as a relay node Wth relatvely rch resources, moble robots servng as relay transmsson tasks can let the normal nodes focus on data acquston. There s no need to transfer data to the convergent nodes by mult-hop. Sensor nodes wll drectly transfer the data to a moble relay node wth one jump, and the moble relay node wll transmt the data to the convergent node. Ths program also can greatly mprove the energy effcency of the sensor networks, because the orgnal long-dstance multhop communcaton becomes a short-range snglehop communcaton Detecton algorthm of R-based V- BLAST transmsson In terms of the performance, the maxmum lkelhood detector (LD) s optmal n the sense of mnmzaton of bt error rate, but t s not practcal for ts complexty. Its complexty ncreases exponentally wth the number of transmsson antennas and the modulaton sze. In terms of transmsson antennas and usng CQA as an example, L detecton can be expressed as: ^ = arg mn r Hx j x m j C x (10) Where C s the number of constellaton. L algorthm s not practcal, so t s necessary to study more effcent algorthms. Snce L detecton algorthm s complexty comes from detectng every layer s possble m symbols n aggregate C, then the queston s that m every element n C has the same probablty? Obvously, t s not every detecton s useful. So, the basc method to reduce complexty of L algorthm s to reduce the number of elements(symbols) n constellaton whch s used to decode[11]. So, we wsh to fnd one transmtter antenna, W (W<C) ponts n the constellaton as a subset to the m L detecton. It means that C tmes detecton s decreased to W (W<C) tmes n the effcent L detecton algorthm. Frstly, we select a neghborhood of the best soluton (the detected symbol of the optmal layer) of the OSIC algorthm as the canddate detecton set. The canddate set of neghborhood s shown n Fg 1. Secondly, we use OSIC algorthm agan to obtan other layers canddate symbols. The specfc method s descrbed partcularly n Fg. 4. Fgure 4: Subset Of Constellaton Pont In Fg. 4, the symbol wth a pentagram sgn n modulaton constellaton dagram s the best solutons of V-BLAST algorthm. Let s place the pentagram at the center and draw a crcle, we get a neghborhood around the strongest layer s detected symbol by OSIC. 680

5 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: As can be seen from Fg. 4, there are two factors that nfluence the number of constellaton ponts around the best soluton s the neghborhood. One s the radus of the crcle, and the other s the locaton of the best soluton n the constellaton dagram. Obvously, there are three constellaton ponts around the best soluton n Fg. 4(a), and four ponts n Fg. 4(b), two ponts n Fg. 4(c). The effcent Approxmate axmum Lkelhood Detecton procedure for IO communcaton systems s depcted n Fg. 5. In ths Fg., 1 W x, x, L, x denotes the one-dmensonal k k k canddates of maxmum lkelhood detecton, n whch W s the number of constellaton ponts n the neghborhood, k s the mark of the strongest layer of V-BLAST algorthm (normally last layer s the best, that s, k=m ). The detecton algorthm n Fg. 5 can be further descrbed as the followng steps: Step 1: Usng the orgnal V-BLAST algorthm to get the sgnal of the strongest (optmal) layer. Step : Choosng a neghborhood around the symbol detected by OSIC of the optmal layer as the canddate set for the new algorthm. Step 3: Choosng a possble symbol xk from the canddate set. Step 4: Cancellng the effect of nterference n receved vector r caused by x k,then get a new vector r, ths step can be expressed as: r = r H 0, Lx k,0, L,0 (11) Step 5: lmnatng the k-th column of channel transfer matrx H H, then get matrx S. Step 6: Accordng to vector r H and matrx S, Use OSIC algorthm to detect these symbols : x% 1, Lx% ( 1), x% ( 1), Lx% k k + m. Step 7: From =1 to W, repeatng step 3 to step 6, we can get a group of symbols as follows: x% 1, Lx% ( 1), x% ( 1), Lx% k k + m x% 1, Lx% ( 1), x% ( 1), Lx% k k + m ; ;... W W W W x% 1, Lx% ( 1), x% ( 1), Lx% k k + m ;. Step 8: Usng axmum lkelhood rule to determne the output x n step 7. Ths detecton rule can be denoted as: ~ ~ ~ ~ P x = arg m n r H x 1, L x ( k 1), x k, x ( k + 1), L, x m = 1,, L, m m Fgure 5: The New Reduced Complexty L Detecton Algorthm Procedure In order to compare, the performance of the tradtonal OSIC detecton algorthm, L detecton algorthm and the proposed reduced complexty approxmate L detecton algorthm n ths paper are shown n Fg.6. Smulaton s about 4*4 IO systems and 16QA constellaton. Fgure 6: Performance Comparson From the smulaton results we can observe that the proposed reduced complexty L detecton algorthm has more than 10dB performance gan, at the symbol error rate of 10-3, than the tradtonal OSIC algorthm. It s also clear that, the performance of the proposed reduced complexty L detecton algorthm s close to that of tradtonal L detecton algorthm, However, the computatonal complexty s much lower than the L algorthm. 681

6 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: NRGY FFICINCY ANALYSIS AND SIULATION Compared wth the normal data gatherng nodes n WSN, the moble robot s energy can be unrestrcted, so ths part s gnored n the followng dscusson. In ths transmsson program, nergy consumpton s manly concentrated on the remote communcaton from the data gatherng node at the transmtter to the recevng nodes and the local communcaton from the collaboratve nodes to the SINK node at the recevng sde. The communcaton dstance n the wreless sensor network s generally a few hundred meters or less. Crcut power consumpton can not be gnored, compared wth transmsson power. Sometmes crcut power may be greater than transmsson power. The total power consumpton along a sgnal path can be dvded nto two man components: the power consumpton of all the power amplfers PPA and the power consumpton of all other crcut blocks P C. The total power consumpton of the power amplfers can be approxmated as[1] k (4 π ) d l N f PPA = (1 + α) Pout = (1 + α) brb GtGr λ (1) ε Where, P out s the transmsson power, α = 1 η wth η beng the dran effcency of the RF power amplfer and ε beng the peak-to-average rato (PAR) that depends on the modulaton scheme and the constellaton sze. For -QA systems, ε =. 1 d s the transmsson dstance, k s the channel path loss exponent, G t and G r are the transmtter and recever antenna gans respectvely, l s the lnk margn compensatng the hardware process varatons and other addtve background nose or nterference, N f s the recever nose fgure, b s the average energy per bt requred for a gven bterror-rate (BR) specfcaton and R b s the system bt rate. Note that the recever nose fgure N f s gven by N f = Nr / N0 where N r s the power spectral densty (PSD) of the total effectve nose at the recever nput and N 0 s the sngle-sded thermal nose PSD at room temperature [1]. The sgnal attenuaton parameter k could usually le n the range -4 for wreless communcatons channels, wth k= correspondng to free space propagaton. The total crcut power consumpton PC can be dvded nto P ct at the transmtter sde and P cr at the recevng sde. we may estmate the power consumpton as P ct = t( P + P mx + P + P syn) (13) DAC P cr = r ( P + P mx + P + P + P + P syn) (14) Where, P DAC flt LNA IFA flr ADC, P mx, P flt, P syn, P LNA, P IFA, P flr and P ADC are the power consumpton values for the D/A converter (DAC), the mxer, the actve flters at the transmtter sde, the frequency syntheszer, the low nose amplfer (LNA), the ntermedate frequency amplfer (IFA), the actve flters at the recever sde and the A/D converter (ADC), respectvely [1]. Assumng that there are t transmtter and the symbol rate of the every transmtter s R s, the effectve bt rate of the system can be gven by R R log bt = t s (15) Accordng to (13) and (14), the total energy consumpton per bt n a -ary QA IO system s PPA + PC bt = Rbt ε (4 π ) ln f k PC = d b + η Gt Grλ Rbt 3 + 1(4 π ) ln f k PC = d 1 b + η Gt Grλ t log Rs (16) As an example, consderng the smulaton condtons n Ref.[1], the man communcaton parameters are lsted n Table 1. Table 1: Communcaton Parameters Parameters values Parameters values P DAC 15.7 mw η 0.35 P ADC 6.7 mw P LNA 0 mw P fl r.5 mw P IFA 3 mw P flt.5 mw GtGr 5 db f c.5 GHz N0 171 dbm/hz P mx 30.3 mw l 40 db P syn 50 mw N f 10 db 4.1 oble Robot Recevng Wth Its Own ult- Antenna In the frst scheme, the moble robot has ts own mult-antenna. Assumng that, the modulaton scheme s 16QA constellaton, t = r = 4,. The system requrement of SR s 0.001, and the channel path loss factor s k =. Assumng that the moble robot has no energy constrants, the total energy consumpton of the VIO communcaton transmsson n sensor 68

7 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: networks can be approxmated as the energy consumpton of the long-haul communcatons at the transmtter sde. So the average energy per bt for VIO communcatons system n the WSNs can be wrtten as: 1(4 ) mmo 3 + π l Nf k Pct b = bt = d (1 1 b + η GG t rλ t log Rs 7) Fgure 7 shows the total energy consumpton of the 16QA constellaton, 4 4 VIO communcaton transmsson n sensor networks for dfferent communcaton dstances. At the transmtter sde, there are four data gatherng nodes sendng total 104 (4 * 56) bts. At the recever sde, the moble robot has four recevng antennas. Fgure 7: Total nergy Consumpton Of 4 4 IO Vs. SISO In The Scheme Of oble Robot Recevng Wth Its Own ult-antenna As an example for 100m transmsson dstance, the energy consumpton s joules n SISO transmsson scheme. Whle n the robot-based 4 4 VIO transmsson scheme, the energy consumpton s decreased to joules and joules, when OSIC algorthm and approxmate L algorthm are adopted respectvely. They are above 51.6% and 4.8% relatve to the SISO transmsson scheme. Obvously, the proposed scheme effectvely reduces network energy consumpton, and the lowcomplexty maxmum lkelhood detecton algorthm further mproves energy effcency. 4. Collaboratve Nodes Assstng oble Robot ult-antenna Recevng Assumng that there are t data gatherng nodes at transmtter sde, and L bts needs to be sent to the snk node for every gatherng node. At the recever sde, the moble robot serves as the SINK node and r 1 nodes around partcpate n collaboratve recevng for the moble robot. After recevng sgnal sample from the transmtter sde, collaboratve nodes send the receved sgnal to the moble robot(snk node) n the next tme slot. The robot makes an approprate choce of the V-BLAST decodng algorthm to obtan the sgnals transmtted by data acquston nodes, accordng to the sgnal receved from ts own antenna and all the collaboratve nodes. The total energy consumpton from the data gatherng nodes to the robot can be calculated as mmo ( L) ( r 1) qlpct ( l) = t L + (17) bt ( ) Where, L bt denotes the average energy consumpton per bt n the long-haul transmsson. ( l ) bt denotes the average energy consumpton per bt n the local transmsson. denotes the QA constellaton sze used for long-haul communcatons and thus L / log s the total number of sgnal samples receved by the sensor nodes at the recever sde. q s bts number needed for the each receved sample, from collaboratve nodes to the robot. Snce durng each tme-slot, only one collaboratve node s communcatng wth the robot at the recever sde. As long-haul communcatons, the total energy per bt for a fxed rate QA system can be estmated as log ( l) PPA + PC l f k ( l ) PC bt = d ( l) b + Rbt η 1 GtGr λ log Rs (4 π ) N = (18) Where crcut power consumpton Pc durng local communcatons can be wrtten as P c = ( P LNA + P mx + P IFA + P flr + P ADC + P syn ) + ( P DAC + P mx + P flt + P syn ) (19) Note that, the moble robot has no energy constrants, P c can be estmated by the second part Pc = ( PDAC + Pmx + Pflt + Psyn ).Accordng to Ref. [1], n case of an AWGN local channel, the average ( ) transmttng energy consumpton per bt l b can be descrbed as N ( 1) BR log ( ) = Q ( ) ( l) 0 1 b 3log ( ) 4(1 (1/ ) bt (0) In case of a Raylegh fadng local channel, the average transmttng energy consumpton per ( ) bt l can be descrbed as b N ( 1) BR log ( ) ( l ) 0 1 b = ((1 ) 1) 3log ( ) (1 (1/ )) (1) Assumng that the local channel s Raylegh fadng, there are three collaboratve nodes around the moble robot wth a sngle antenna. The communcaton dstance from collaboratve nodes to the robot s 10m and the average error rate s At the transmtter sde, usng 16QA constellaton, there are four data gatherng nodes transmttng 104 bts respectvely. 683

8 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: Fgure 8 shows the energy consumpton when collaboratve nodes assstng moble robot multantenna recevng scheme s adopted. From Fgure 7 and Fgure 8, we can see that the total energy consumpton of ths scheme s hgher than that of moble robot recevng wth ts own mult-antenna. That s because the dscusson here s based on the assumpton, the energy consumpton of the robot has no constrants, and some collaboratve nodes at the recever sde partcpate n the mult-antenna recevng of the robot. Fgure 8: nergy Consumpton Of Collaboratve Nodes Asssted oble Robot ult-antenna Recevng As an example for 100m transmsson dstance, the energy consumpton s joules when OSIC algorthm was adopted. Whle the energy consumpton s decreased to joules, when approxmate L algorthm was adopted, whch s only 1% the energy consupton of that n the former. Relatve to OSIC algorthm, approxmate L detecton algorthm proposed n ths paper further mproves energy effcency. As Ref.[1], we can defne energy SISO IO effcencyη =.Where, IO s the energy SISO consumpton n collaboratve node assstng moble robot mult-antenna recevng scheme, SISO s the energy consumpton of transmttng the same number of bts, when SISO communcaton transmsson scheme, the 16QA constellaton s adopted. Fgure 9 shows the energy effcency curves for dfferent channel path loss factor k, n collaboratve node assstng moble robot mult-antenna recevng scheme. The detecton algorthm of robot s based on the approxmate L. Fgure 9: nergy ffcency Of Collaboratve Nodes Assstng oble Robot ult-antenna Recevng It can be seen from Fgure 9, wth the ncreases of transmsson dstance, the energy effcency s hgher, and the energy savng effect s more obvous. It can be seen from Fgure 9, wth the ncreases of path loss factor k, the energy savng effect ncreases markedly n the area where transmsson dstance s relatvely small. It means that transmsson energy consumpton of the collaboratve transmsson of the asssted node s domnant n the total communcaton energy consumpton, especally for large values of k. 5. CONCLUSION 1) Amng at the requrements of hgh-speed data transmsson and energy consumpton n the case of emergency rescue and real-tme detecton, a moble robot-based vrtual V-BLAST IO transmsson scheme n dstrbuted WSNs s proposed. oblty of robot can reduce the communcaton dstance between the sensor nodes, IO-based WSN can use ts dversty gan to overcome fadng effects and can also use ts multplexng gan to ncrease the data transmsson rate. Smulaton results show that the proposed scheme effectvely reduces network energy consumpton. ) Amng at the moble robot-based vrtual V- BLAST IO transmsson scheme, approxmate maxmum lkelhood algorthm s proposed n ths paper. Combned wth tradtonal decodng OSIC algorthm, the vectors need to be detected s reduced, and the computatonal complexty s decreased greatly. Smulaton results show that the low-complexty approxmate maxmum lkelhood detecton algorthm further mproves energy effcency, relatve to tradtonal OSIC algorthm. ACKNOWLDGNTS Ths work was supported by a Fundng Project for Academc Human Resources Development n Insttutons of Hgher Learnng under the 684

9 Journal of Theoretcal and Appled Informaton Technology JATIT & LLS. All rghts reserved. ISSN: ISSN: Jursdcton of Bejng uncpalty (NO. PHR , PHR ). RFRNCS: [1] Shuguang Cu, A. J. Goldsmth, A. Baha, nergy-effcency of IO and cooperatve IO technques n sensor networks, I Journal on Selected Areas n Communcatons, Vol., No. 6, 004, pp [] Je Zhao,Gangfeng Lu, ult-robot system for search and exploraton n the underground mne dsasters based on WSN, Journal of Chna Coal Socety, Vol. 34, No.7, 009, pp [3] G. J. Foschn,. J. Gans. On lmts of wreless communcatons n a fadng envronment when usng multple antennas, Wreless Personal Communcatons, No.6, 1998, pp [4] J. N. Laneman, G. W. Wornell, Dstrbuted space-tme-coded protocols for explotng cooperatve dversty n wreless networks, I Transactons on Informaton Theory, Vol. 49, No. 10, 003, pp [5]. Grossglauser, D. Tse, oblty ncreases the capacty of ad-hoc wreless networks, I/AC Transactons on Networkng, Vol. 10, No. 4, 00, pp [6] Wench Cheng, Haln Zhang, Approxmatng maxmum lkelhood performance reduced dmenson VBLAST detecton algorthm, Scence Chna (Informaton Scences), No. 7, 010, pp [7] Hahong Wang, Xn Wang, Jbo We, Group axmum Lkelhood Detecton Algorthm for a 4 4 V-BLAST System. Sgnal Processng, Vol. 6, No. 3, 010, pp [8] Xaobe L, Jelng Wang, Yongshun Zhang, Jont Detecton Algorthm for V-BLAST System wth Hgher Performance, Journal of System Smulaton, Vol. 1, No. 5, 009, pp [9] T. Cover, A.. Gamal, Capacty theorems for the relay channel. I Transactons on Informaton. Theory, Vol. 5, No. 5, 1979, pp [10] A. Sendonars,. rkp, B. zhang, Increasng uplnk capacty va user cooperaton dversty, Proceedngs of I Internatonal Symposum on Informaton Theory, I Informaton Theory Socety, August 16-1, 1998, pp.156 [11] Wenba Chen, We L, Xaopn Zhang, Complexty reducton L Detecton algorthm for IO system, Journal of Harbn Insttute of Technology, Vol. 44, No. 5, 01, pp [1] S. K. Jayaweera, V-BLAST-Based Vrtual IO for Dstrbuted Wreless Sensor Networks, I Transactons on Communcatons, Vol. 55, No. 10, 007, pp