World Academy of Scence, Engneerng and Technology 5 2 An Effcent Energy Adaptve Hybrd Error Correcton Technque for Underwater Wreless Sensor Networks Ammar Elyas babker, M.Nordn B. Zakara, Hassan Yosf, and Samr B. Ibrahm Abstract Varable channel condtons n underwater networks, and varable dstances between sensors due to water current, leads to varable bt error rate (BER). Ths varablty n BER has great effects on energy effcency of error correcton technques used. In ths paper an effcent energy adaptve hybrd error correcton technque () s proposed. adaptvely changes error technque from pure retransmsson () n a low BER case to a hybrd technque wth varable encodng rates ( & FEC) n a hgh BER cases. An adaptaton algorthm depends on a precalculated packet acceptance rate (PAR) look-up table, current BER, packet sze and error correcton technque used s proposed. Based on ths adaptaton algorthm a perodcally 3-bt feedback s added to the acknowledgment packet to state whch error correcton technque s sutable for the current channel condtons and dstance. Comparatve studes were done between ths technque and other technques, and the results show that s more energy effcent and has hgh probablty of success than all those technques. Keywords Underwater communcaton, wreless sensor networks, error correcton technque, energy effcency I. INTRODUCTION NDERWATER wreless sensor networks fnd many U applcatons n oceanographc data collecton, envronmental montorng, dsaster preventon and ol exploraton []-[2]-[3]. For such applcatons a relable and effcent communcaton data transport s demanded [4]. and FEC are the two man error correcton technques that guarantee relablty n underwater envronment [5]. Energy s the most mportant effcency ssue for underwater wreless sensors due to the dffculty n rechargng or replacng batteres n most aquatc medum [4]. In ths paper an effcent energy adaptve hybrd error correcton technque () s proposed. Ths technque depends on a proposed adaptaton algorthm whch s calculated based on a pre-calculated packet acceptance rate (PAR) look-up table, current bt error rate (BER), packet sze, and last error correcton technque used. Based on ths adaptaton algorthm output a perodcally 3-bt feedback s added to the acknowledgement packet to Frst and second authors are wthn computer and nformaton scence department, Unverst teknolog PETRONAS, 375, tronoh, perak, malaysa, e-mal: ammaralyas@gmal.com). * Thrd and fourth authors are wth electrocal and electronc engneerng department, Unverst teknolog PETRONAS, 375, tronoh, perak, Malaysa. state whch error correcton technque s sutable for the current channel condtons and dstance. The error correcton s chosen from a pure n a good channel condtons and short dstances to an wth varable encodng rates n bad channel condtons and long dstances.in [6], we have performed an energy effcency analyss to and FEC n underwater envronment. Energy effcency of both technques depends on channel condtons, transmsson power, dstance, and packet sze. s found to be more energy effcent n some cases, whle FEC s more effcent n others. In [7] a propagaton model to calculate the sgnal to nose rato for underwater acoustc channel was desgned and mplemented. In [8] modulaton and encodng technques for underwater communcaton system were studed. It was found that 8-PSK s the best modulaton for underwater systems, whereas convoluton codng s found to acheve better codng gan so t s the modulaton and encodng technques used n ths work. In [9] an optmzaton metrc for energy effcency was proposed, and t was used n [6, ] for energy effcency calculatons. In [] Tan et el. have proven that energy effcency of technques s ndependent of retransmsson attempts,; they compared and FEC technques for terrestrals wreless sensor networks n terms of energy effcency.in [] s proposed for mult-hop underwater communcaton channel, the acknowledged can be acheved by explctly transmttng the ack., or mplctly by hearng that packet transmtted forward to the next hop.in [2] an effcent s proposed by utlzng the sharng propertes of underwater channel (schedulng packets transmsson to acheve collson free transmsson). In [3] jugglng concepts enable a contnuous rrespectve of half-duplex propertes of acoustc. Ths leads to hgh throughput, but not affectng energy effcency. In [4] an opportunstc mult-hop s mplemented n real system. Ths provdes mprovements n terms of data delvery rato, but end to end delay ncreases due to queung and retransmsson. In [5], network codng technque s proposed to make use of the broadcast nature of acoustc channel. It s good n error recovery, but at the cost of energy effcency. In [6], ARRTP (Adaptve Redundancy Relable Transport Protocol), prevously known as ADELIN [7] (an ADaptve relable transport protocol) s proposed. Three schemes whch combne forward error correcton (FEC) mechansm at the bt and/or packet level n non-cooperatve and cooperatve scenaros were proposed. ARRTP uses dfferent schemes for 389
World Academy of Scence, Engneerng and Technology 5 2 dfferent dstances dependng on trade-off between relablty and energy consumpton. In ths paper an effcent energy for underwater communcaton system s proposed. A comparson between ths technque and the technques that use pure and FEC s done. s compared wth ARRTP after an energy effcency analyss for scheme and 3 were done. ARRTP scheme 2 s neglected as t s found to be neffcent [6]. It s also compared wth the system when varable power supply s used as adaptaton factor.due to space lmtaton, we are not gong to repeat the underwater propagaton model and the mathematcal energy effcency analyss for and FEC whch s found n [6]. Our contrbuton can be summarzed as follows: Energy effcent Adaptve Hybrd Error Correcton Technque s proposed for UWSN. An adaptaton algorthm whch depends on a Precalculated PAR ranges look-up table, current BER, packet length, and current error correcton technque s proposed, ths algorthm adapt to the varaton n both channel condtons and dstances. In secton 2 we present the man dea of our proposed. The adaptaton algorthm wll be presented n secton 3. In secton 4 we wll present how to calculate the look-up table. Adaptaton usng varable power supply wll be presented n secton 5. Energy effcency analyss for ARRTP wll be presented n secton 6. Results and analyss n secton 7, and n secton 8 the paper s concluded. II. MECHANISM The results of the analyss n [6] state that energy effcency of error correcton technques vares wth the varaton n transmsson dstances and channel condtons. In some cases, one technque s better than the other, and vce versa. Wth ths n mnd we propose whch acheves hgh energy effcency n a varyng dstance, varable channel condton cases by adaptvely changes the error correcton technque used.the technque works lke ths: for varable dstances and varable channel condtons, always search for the technque wth the hghest energy effcency, and snce relablty s one part n energy effcency calculaton, t wll also be a relable technque. The technque depends on an adaptaton algorthm whch based on the current packet acceptance rate (PAR), current error correcton technque used, and a pre-calculated PAR ranges look-up table to determne whch error correcton technque s sutable for the current dstance and current channel condtons. In, only modulaton technque (.e. ) s used n good channel condtons and short dstances, whch means low BER. Selectve repeat wll be the most sutable type of for two reasons: As the BER s very low, no acknowledgement s needed for every packet, so ether Go-Back-N or selectve repeat s most sutable. In Go-back-N, the error packet and all the subsequent packets wll be retransmtted, whch results n a waste of energy; whereas n selectve repeat only the error packet wll be retransmtted. In bad channel condtons and long dstances varable code rates convolutonal codng are used. Convolutonal code s used for two reasons also: It s the best encodng technque for underwater communcatons as stated by [8]. Wth convolutonal codng, we can easly use puncturng technque to obtan varable code rates, whch s needed n our. Varable code rates are obtaned usng puncturng technque by deletng part of the bts of low-rate convoluton code [8] as n the Table, and t s represented n MATLAB usng systematc puncturng convoluton codes wth the parameters obtaned from [8] as shown n Table II.: III. ADAPTATION ALGORITHM TABLE I PUNCTURING MATRIX Code rate Puncturng Matrx 2/3 [ ] 3/4 [ ] 4/5 [ ] 5/6 [ ] 6/7 [ ] adaptaton algorthm can be descrbed as follows: Usng error detecton technque n the recever, BER s perodcally calculated, and from whch PAR s calculated usng the packet length n as: PAR = ( BER) n () Then the sutable error correcton technque s calculated from the functon: J = f ( PAR, I, PARMAX ( I, J ), PARMIN( I, J )) (2) Where J s the sutable error correcton technque requred, PAR s the current packet acceptance rate, I s the current error correcton technque used, and PARMAX ( I, J ), PARMIN( I, J ) are the maxmum and mnmum values n the pre- calculated PAR lookup table ranges. We can mathematcally model ths functon as n the followng formula: 6 J = n I ( PAR) (3) n= Where s the values n the look-up table taken from the energy effcency analyss of sx error correcton technques (One and fve varyng code rate FEC) [6], and (4) From the value of J obtaned, a 3-bt feedback s added to the acknowledgement to state whch error correcton technque to use as n table (3) A I n 39
World Academy of Scence, Engneerng and Technology 5 2 TABLE III ERROR CORRECTION TECHNIQUES DETAILS Error correcton Conssts of FEC Code Feedback technque Rate Pure 2 Hybrd & FEC 6/7 3 Hybrd & FEC 5/6 4 Hybrd & FEC 4/5 5 Hybrd & FEC 3/4 6 Hybrd & FEC 2/3 The adaptaton algorthm can be wrtten as n algorthm Fg. below: Adaptaton Algorthm Feed Back ( BER current (current BER), n (packet length), PARMAX(I,J), PARMIN(I,J), I (current error correcton technque)) let J= n 2 PAR current = ( BER current ) 3 If PARMIN(I,J) < PAR current < PARMIN(I,J) 4 Sutable Error Correcton Technque = J 5 Go to 8 6 else J=J+ 7 Go back 3 8 If J =, then Feed Back = 9 Else If J=2, then Feed Back = Else If J = 3, then Feed Back = Else If J = 4, then Feed Back = 2 Else If J =5, then Feed Back. = 3 Else Feed Back. = 4 end Return (Feed Back) Fg. Adaptaton algorthm IV. PRE-CALCULATED LOOKUP TABLE CALCULATIONS The pre-calculated lookup table s calculated as follows:. Energy effcences and PARs of the sx error correcton technques ( plus fve varable code rates FECs) for varable values of SNR are found as n [6], (SNR are taken as a measure for dstance and channel condtons varatons). 2. Startng wth the SNR values whch gves PAR values equal to for all the technques; at ths SNR wll have the maxmum energy effcency compared to the others, so the PAR for all those technque at ths pont s the maxmum values n the ranges whch makes the sutable technque s technque (pure ). Ths means PARMAX J, =,.e. f the current technque s J and the current PAR s n the range that has as the maxmum value, then technque one s the most energy effcent technque. 3. Then decreasng SNR value untl the energy effcency of the frst technque s less than the energy effcency of the second technque; at ths SNR the PAR for all technques wll be the mnmum values n the ranges whch makes the sutable technque s technque (pure ). Ths means the PAR of any technque J at ths pont = PARMIN J,.e. f the PAR of the current technque J s n between PARMIN J, and PARMAX J,, then technque s the most energy effcent technque. As the mnmum values n the frst range equal the maxmum values n the second range, then: PARMAX J,2 = PARMIN J, 4. Repeat step three above and ncreasng one to the error correcton technque n each tme untl we come to the last technque as n the algorthm Fg. 2 below: Look-up table calculaton algorthm PARMAX(I,J), PARMIN(I,J)(E.Ef (J, SNR) from [], ( Energy Effcency calculaton), and (FEC Energy Effcency calculatons)) for J = :6; 2 PARMAX J, =; PARMIN J, 6 = ; 3 SNR = SNRMAX; 4 For I= :5; 5 If E.Ef (I, SNR)<=E.EF (I+, SNR); 6 then PARMIN J, I = PAR (J, SNR); 7 PARMAX J, I + = PAR (J, SNR); 8 SNR = SNR -; 9 else go to 5; end; end; 2 return ( PARMIN J, I, PARMAX J, I )//The maxmum and mnmum values n the lookup table from error correcton technques to error correcton technques 6// Fg. 2 Look-up table calculaton algorthm V. ADAPTIVE VARIABLE POWER SUPPLY (AVPS) Adaptvty can also be acheved usng varable power supply. For dfferent channel condtons and dfferent dstances between sensor nodes varable transmt power values can be used to acheve the hghest energy effcences usng the same dea of adaptaton algorthm. When usng varable power supply as adaptaton, wth sx dfferent power supply values as n Table 4 s used nstead of the sx error correcton technques used n to calculate the pre-calculated look-up table. The energy effcency n case of varable power supply can be calculated usng the followng formula: eff E Pref Eff vps = ( PER tot ) E P Where l Pref = ( PER ) (5) l + α + τ + ack Pt Eff s the energy effcency for when vps usng varable power supply, t P ref s a reference transmt power, or the desgned power, P t s the varable transmt power, and ( PER, l, α, τ, ack ) from [6]. 39
World Academy of Scence, Engneerng and Technology 5 2 From the pre-calculated lookup table, current PAR, and current power supply value used, the sutable power supply value whch wll gves the most energy effcent transmsson can be calculated. TABLE IV AVPS ERROR CORRECTION TECHNIQUES DETAILS Error Correcton Transmt power (Watt) Feedback technque 2. 2 2.5 3 3. 4 3.5 5 4. 6 4.5 BCH RS Check BCH Check VI. INTERNODES DISTANCE-BASED REDUNDANCY RELIABLE TRANSPORT PROTOCOL (ARRTP) ENERGY EFFICIENCY ANALYSIS Energy Effcences for the dfferent ARRTP schemes, [6], Fg. 3, prevously known as ADELIN [7] s found as follows: For non-cooperatve scheme-, t n n PAR scheme = ( Pb ) ( Pb ) (6) = Where PAR scheme s the packet acceptance rate, t s the correctablty factor, n s the packet length. Eff eff Escheme scheme = tot scheme Escheme ( PAR ) n α φ = ( PARscheme n ) (7) For non-cooperatve scheme-3, snce reconstructng k orgnal data packets needs recevng any k packets out of (k+s) packets, the probablty of successfully transmsson of k packets s gven by: k + s k + s K + s PAR scheme 3 = ( Ps ) ( Ps ) = k (8) Where P s s the probablty of successfully transmttng one packet over one hob wth BCH codng for scheme-3 whch s the same as equaton (2). For scheme-3, energy effcency can be gven by: (9) Where φ s overhead n the data packet due to BCH codng φ2 s the overhead n the check packet due to BCH codng. Scheme-2 s gnored as t s neffcent compared wth scheme- and scheme-3 [6]. (a) Scheme- (BCH) (b) Scheme-3 (BCH-RS) Fg. 3 Dfferent ARRTP schemes VII. RESULTS AND DISCUSSION A. Versus and FEC Probablty of Success and Energy Effcency From Fg. 4 below; t s clear that has hgher probablty of success (PAR) compared wth both and FEC, except for a short dstance from 22 m to 24 m; where FEC has a hgher probablty of success than. Ths dfferences whch s around 2-3 % has no notceable effect on the system snce both technques have more than 9 % probablty of success. PAR.9.8.7.6.5.4.3.2 Vs & FEC Probablty of Success. FEC (C.R. =5/6) 5 5 2 25 3 Fg. 4 Vs & FEC Probablty of Success Fg. 5 gves a comparson between the energy effcency of and pure and FEC for varyng dstances. From ths fgure t s clear that s more energy effcent than both and FEC n varable dstances stuaton. Compared wth the pure, acheves % ncrease n savng energy when the dstance s around 5 m to more than 6 % when the dstance ncreases above 7 m. When compared wth FEC, t acheves around % ncrease n energy savng when the dstance s below 5 m, and around 7 % savng when the dstance goes above 5 m. between 24 and 26 m, both and FEC have the 392
World Academy of Scence, Engneerng and Technology 5 2 same energy effcences as uses the same code rate of ths FEC..8 Vs FEC & Energy Effcences short dstances (around m) to more than 8 % n long dstances (around 24 m). Vs AVPS Proablty of Success.7.9.6.8.7.5.4.3 PAR.6.5.4.2.3. FEC (Code Rate = 4/5) Pure 5 5 2 25 3 Fg. 5 Vs & FEC Energy Effcency (Varable Dstances Case) In Fg. 6 varable wnd speed s taken as a measure for the varaton n channel condtons. From ths fgure t s clear that s more energy effcent than both and FEC for varable wnd speed (.e. varable channel condtons). Compared wth the pure, and when the transmsson dstance s 5 m, acheves 5 % ncrease n energy savng when there s wnd of.5 m/s speed, more than 6 % energy savng when wnd speed ncreases to m/s. When compared wth FEC, acheves around 7 % ncrease n energy savng when there s no wnd, and around 5 % when the speed s greater than.5 m/s..8.7.6 Vs FEC & (Varable Channel Condtons) FEC.2. AVPS 5 5 2 25 3 Fg. 7 Vs AVPS probablty of Success Fg 8 below gves a comparson between energy effcency of and AVPS for varyng dstances. From ths fgure t s clear that s more energy effcent than AVPS n varable dstances stuaton. Compared wth AVPS, acheves 2 % ncrease n savng energy when the dstance s around 7 m, and more than 6 % when the dstance ncrease above 22 m..8.7.6.5.4.3 Vs AVPS Energy Effcency (varable Dstance).5.4.3.2..5.5 2 2.5 3 Wnd Speed (m/s) Fg. 6 Vs & FEC Energy Effcency (Varable Channel condtons) B. Versus AVPS Probablty of Success and Energy Effcency From Fg. 7, t s clear that has hgher probablty of success than AVPS. Ths dfferences range from % n.2. AVPS 5 5 2 25 3 Fg. 8 Vs AVPS Energy effcency (Varable Dstance Case) In Fg. 9 varable wnd speed s taken as a measure for the varaton n channel condtons. From ths fgure t s clear that s more energy effcent than AVPS for varable wnd speed (.e. varable channel condtons). Compared wth AVPS when the transmsson dstance s 5 m, acheves 5 % ncrease n savng energy when there s wnd of speed.5 m/s, more than 3 % when wnd speed ncreases to m/s, and more than 6 % when wnd speed s greater than 2 m/s. 393
World Academy of Scence, Engneerng and Technology 5 2.8 Vs AVPS (Varable Channel Condtons).8 Vs ARRTP Energy Effcency (Varable Dstance Case) ARRTP(BCH alone) ARRTP (BCH & CS).7.7.6.6.5.5.4.3.4.3.2.2...5.5 2 2.5 3 Wnd Speed (m/s) Fg. 9 Vs AVPS Energy Effcency (Varable Channel Condtons) C. Versus ARRTP Probablty of Success and Energy Effcency From Fg. 8 below; t s clear that has hgher probablty of success than ARRTP. Ths dfferences range from around 2 % when the transmsson dstance s around 2 m to more than 9 % when the transmsson reach 25 m. t s also clear that ARRTP wth both BCH and CS has hgher probablty of success than when only BCH codng s used. PAR.9.8.7.6.5.4.3.2 Vs ARRTP Probablty of Success 5 5 2 25 3 Fg. Vs ARRTP Energy Effcency (Varable Dstances Case) Regardng varable channel condtons, ARRTP don t adapt to the varatons n channel condtons, as t uses approxmate formula for calculatng noses, whch gnore the effect of wnd speed and shppng factors. VIII. CONCLUSIONS In ths paper, we have presented our dea, and how the adaptaton occurs. The adaptaton algorthm s based on: A pre-calculated PAR look-up table whch s calculated from the energy effcency analyss we have done on [6]. Current BER, whch can be easly determned usng any error detecton technques, packet length, and current error correcton technques. Based on the results of the adaptaton algorthm the recever sends 3-bt feedback wth the acknowledgement tellng the sender whch error correcton technque s most sutable for the current dstance and current channel condtons. The results show that our has more probablty of success and more energy effcent than all the other error correcton technques. ARRTP (BCH only). ARRTP (BCH & CS) 5 5 2 25 3 Fg. Vs ARRTP probablty of Success Fg. below gves a comparson between the energy effcency of and ARRTP for varyng dstances. From ths fgure t s clear that s more energy effcent than both types of ARRTP n varable dstances stuaton. Compared wth ARRTP (Only BCH), acheves 5 % ncrease n energy savng when the dstance s below 5 m, and more than 5 % when the dstance ncrease above 22 m. when compared wth ARRTP (BCH and CS), acheves around 25 % savng n energy when the dstance s less than 7 m and more than 5 % when the dstance s more than 22 m. REFERENCES [] J. Hedemann, et al, Research Challenges and Applcatons for Underwater Sensor Networkng, Proceedngs of Wreless Communcatons and Networkng Conference., 228-235, 26. [2] G. John, et al, Shallow Water Acoustc Networks, Communcatons Magazne, IEEE, vol. 39, no., pp. 4 9, 2. [3] I. F. Akyldz, et al, Challenges for Effcent Communcaton n Underwater Acoustc Sensor Networks, ACM Sgbed Revew, vol., no. 2, 24. [4] P. Xe, et al, An FEC-based Relable Transport Protocol for Underwater Sensor Networks, Proceedngs of 6th Internatonal Conference on Computer Communcatons and Networks, ICCCN7, pp.747-753, 27. [5] I.F. Akyldz, et al, State-of-the-Art n Protocol Research for Underwater Acoustc Sensor Networks, Proceedngs of the6th nternatonal workshop on underwater networks, 26. [6] Ammar Elyas Babker, M. Nordn B. Zakara, Energy Effcency Analyss of Error Correcton Technques n Underwater Wreless Sensor Networks. Journal of Engneerng Scence and Technology (JESTEC), vol. 6, no., pp. 7-28, 2. 394
World Academy of Scence, Engneerng and Technology 5 2 [7] A. F. Harrs, M. Zorz, Modelng the underwater acoustc channel n ns2, Proceedngs of the 2nd nternatonal conference on Performance evaluaton methodologes and tools, Vol. 32, no. 8, 27. [8] Y. Labrador, et al, Modulaton and Error Correcton n the Underwater Acoustc Communcaton Channel, IJCSNS Internatonal Journal of computer Scence and Network Securty, Vol. 9, No. 7, pp. 23-3, 29. [9] Y. Sankarasubramanam, et al, Energy Effcency based Packet Sze Optmzaton n Wreless Sensor Networks,. Proceedngs of the Frst IEEE Internatonal Workshop on Sensor Network Protocols and Applcatons, 23. [] Z. Tan, et al, Energy Effcency Analyss of Error Control Schemes n Wreless Sensor Networks, Proceedngs of Internatonal Wreless Communcatons and Moble Computng Conference, IWCMC '8, pp. 4-45, 28. [] H.-P. Tan, W. K. G. Seah, and L. Doyle, A mult-hop arq protocol for Underwater Acoustc Networks. [2] J.-W. Lee, J.-P. Km, J.-H. Lee, Y. Jang, K. C. Dho, K. Son, and H.-S. Cho, "An mproved scheme n underwater acoustc sensor networks," n Proc. MTS/IEEE Oceans, Kobe, Japan, Apr. 28, pp.-5 [3] Mngsheng Gao, Wee-Seng Soh and Mexa Tao, "A Transmsson Scheme for Contnuous Protocols over Underwater Acoustc Channels," IEEE ICC, Dresden, Germany, Jun. 29. [4] Alvn C. Valera, Pus W. Q. Lee, Hwee-Pnk Tan, H. G. Lang and Wnston K. G. Seah. Implementaton and Evaluaton of Multhop for Relable Communcatons n Underwater Acoustc Networks. Accepted for publcaton at IEEE Oceans, May 29 [5] Zheng Guo, Peng Xe, Jun-Hong Cu, Bng Wang, On applyng network codng to underwater sensor networks, Proceedngs of the st ACM nternatonal workshop on Underwater networks, September 25-25, 26, Los Angeles, CA, USA [6] L. Bn, et al, Inter-node Dstance-Based Redundancy Relable Transport n Underwater Sensor Networks. EURASIP Journal on Wreless Communcatons and Networkng, Volume 2, Artcle ID 3587, 2. [7] L. Bn, et al, A study of forward error correcton schemes for relable transport n underwater sensor networks, Proceedngs of the 5th Annual IEEE Communcatons Socety Conference on Sensor, Mesh and Ad Hoc Communcatons and Networks (SECON '8). pp. 97 25, 28. [8] G. Begn, et al, Further Results on Hgh - Rate Puncturng Convolutonal codes for vterb and sequental decodng, IEEE transacton on Communcatons. Vol. 38, No., pp. 922-928, 99. TABLE II MINIMUM HAMMING DISTANCES ( d free ) AND WEIGHT DISTRIBUTION ( w dfree ) FOR VARIABLE RATE CONVOLUTIONAL CODES 2/3 3 54 226 853 338 432 34836 497 3/4 3 5 4 54 252 48 4656 9448 763944 36844 4/5 2 36 39 258 23 65754 344656 7553 875428 5/6 2 2 974 685 43598 26367 536563 8724988 468477 6/7 2 5 86 942 6423 24469 88752 6889 466478 2662532 395