software-defined underwater acoustic networks: …gsklivan/underwatermag15.pdf · software-defined...

IEEE Communications Magazine • November 201564 0163-6804/15/25.00 © 2015 IEEE

Emrecan Demirors andTommaso Melodia arewith Northeastern Univer-sity.

George Sklivanitis, StellaN. Batalama, and Dim-itris A. Pados are with theState University of NewYork at Buffalo.

ABSTRACT

We review and discuss the challenges of adopt-ing software-defined radio principles in under-water acoustic networks, and propose asoftware-defined acoustic modem prototypebased on commercial off-the-shelf components.We first review current SDR-based architecturesfor underwater acoustic communications. Thenwe describe the architecture of a new software-defined acoustic modem prototype, and provideperformance evaluation results in both indoor(water tank) and outdoor (lake) environments.We present three experimental testbed scenariosthat demonstrate the real-time reconfigurablecapabilities of the proposed prototype and showthat it exhibits favorable characteristics towardspectrally efficient cognitive underwater net-works, and high data rate underwater acousticlinks. Finally, we discuss open research chal-lenges for the implementation of next-generationsoftware-defined underwater acoustic networks.

INTRODUCTIONUnderwater acoustic networks (UANs) are anemerging research topic because of the key rolethat this technology will play in military and com-mercial applications including disaster prevention,tactical surveillance, offshore exploration, pollu-tion monitoring, and oceanographic data collec-tion. A key challenge in the design of UANsstems from the characteristics of the underwateracoustic (UW-A) channel, which exhibits highpath loss, noise, multipath, high and variablepropagation delay, and Doppler spread. There-fore, reliable communication links are practicallyfeasible only at low data rates. Additionally, thepropagation challenges in the underwater envi-ronment result in temporally and spatially varyingUW-A channel coefficients, which drives researchefforts toward the design of specialized protocolsat different layers of the network protocol stack— often with a cross-layer approach.

As of today, the majority of existing deployedUANs are based on commercially available

acoustic modems. Even though commercialmodems enable a wide range of applications,they rely on inherently fixed hardware designs andproprietary protocol solutions that are regrettablyfar from satisfying the emerging reconfigurabilityneeds of next-generation UANs. As a result,practical deployment of new protocols is eithernot feasible or prohibitive in terms of both imple-mentation cost and time. Evidently, fixed hard-ware and closed software architectures thatcharacterize commercial underwater modemsprevent UAN applications from benefiting fromthe latest algorithmic developments.

Lack of standardization agreements forUANs impose additional hurdles in the design ofreconfigurable networks. For example, differentvendors equip underwater modems with propri-etary communication protocols with differentimplementation requirements across differenthardware and software platforms, which, in theend, prevents their integration in heterogeneousUANs. A sizable body of research work in UANsfocuses on the development of software/protocolstandards that resolve interoperability issuesamong different underwater modems (e.g.,JANUS [1]). However, existing architectures arestill far from being able to achieve the data rateperformance and flexibility required by the nextgeneration of UAN applications.

Finally, the spatial and temporal variations ofthe UW-A channel require reconfiguration of thecommunication parameters of UAN devices toprovide stable performance in terms of bit errorrate (BER) and packet error rate (PER). Cur-rently, commercial modems address adaptation tochannel variations by employing a set of prede-fined operational modes (pre-fixed set of commu-nication parameter values). However, such ad hocsolutions lack i) the ability to switch in real timeamong a finite number of operational modes, ii)decision making mechanisms to decide and applyadaptation, and iii) the ability to dynamicallyadapt to all the possible environments because ofthe finite number of operational modes. Conse-quently, there is a need for UAN devices that cani) ease deployment and testing of new protocol

UNDERWATER WIRELESS COMMUNICATIONS ANDNETWORKS: THEORY AND APPLICATION

Emrecan Demirors, George Sklivanitis, Tommaso Melodia, Stella N. Batalama, and Dimitris A. Pados

Software-Defined Underwater AcousticNetworks: Toward a High-Rate Real-Time Reconfigurable Modem

DEMIRORS_LAYOUT.qxp_Author Layout 10/30/15 2:54 PM Page 64

IEEE Communications Magazine • November 2015 65

designs, ii) bridge the gap between different net-work devices and protocols to resolve the interop-erability problem in heterogeneous UANs, andiii) intelligently decide and adapt their communi-cation parameters or technology based on theenvironmental needs in real time.

Software-defined radio (SDR) has recentlyemerged as a technology platform that enablesrapid prototyping of fully agile, intelligently adap-tive, and reconfigurable networking devices toaccommodate and test novel wireless networkingprotocols for RF communications. Consideringthe unique capabilities and features of SDR in RFand the need for flexible, easily reconfigurableUAN devices, in this article, we investigate thesoftware/hardware challenges to build a software-defined acoustic modem (SDAM) and discuss itspotential benefits for future UANs. To that end,we review existing software-defined efforts in UW-A and we propose an SDAM prototype based oncommercial off-the-shelf (COTS) components. Wedesign and implement physical layer schemes anddecision algorithms that exploit both time and fre-quency degrees of freedom, and we demonstratethe real-time reconfigurable capabilities of theproposed modem with field experiments.

The remainder of this article is organized asfollows. We first provide a brief overview of theSDR paradigm and current SDR-based architec-tures for UW-A communications. Then we pro-pose and describe an SDAM architecture andprovide experimental performance evaluationresults in both indoor (water test tank) and out-door (lake) environments. Finally, we discussopen research implementation challenges forfuture next-generation software-defined UANs.

SOFTWARE-DEFINED ARCHITECTURESFOR UW-A COMMUNICATIONS

SDRs are now prevalent communication plat-forms in both commercial and military RF appli-cations mainly because of the need for flexibleand reconfigurable radio solutions and their abil-ity to follow the rapid evolution of enablingtechnologies such as analog-to-digital converters(ADCs), digital-to-analog converters (DACs),general-purpose processors (GPPs), field-pro-grammable gate arrays (FPGAs), and graphicalprocessing units (GPUs). However, so far, SDRplatforms have seen limited use in the context ofunderwater communications, primarily becauseof implementation and development cost chal-lenges compared to RF technologies. A typicalSDR architecture comprises a front-end thatmay include amplifiers, filters, and mixers, con-nected to an ADC/DAC and a softwarerepro-grammable digital processing unit such as adigital signal processor (DSP), FPGA, GPP, orGPU. The SDR architecture provides the basisfor software reconfigurable processing at thephysical (PHY) and medium access control(MAC) layers, often through existing softwaretools, including GNU Radio, Simulink, and Lab-View, which are interoperable with higher-layerprotocol implementations (e.g., TCP/IP).

SDR platforms have been adopted by severalmilitary programs including the U.S. Joint Tacti-cal Radio System (JTRS) and NATO STANAG

5066, as well as by a variety of commerciallyavailable products that feature a mixture of hard-ware configurations at different cost, for exam-ple, USRP, RTL-SDR, HackRF, BladeRF, andPicoSDR. Open-source software packages, suchas GNU Radio interface well with SDR systems,allow custom development of application-specificsignal processing blocks, and have recentlydemonstrated their support for embedded devicestoo such as Xilinx Zedboard and Xilinx ZC702.As a result, a software-defined frameworkappears to be a powerful proposition for under-water communications that will be able to pro-vide a low-cost and programmable/reconfigurablehardware alternative to commercially availableunderwater modems. However, SDR principlesare not directly applicable to the underwaterenvironment due to the particular (and challeng-ing) characteristics of the underwater channel.

Current underwater networking research canbe classified as follows:• Efforts that involve hardware development

of new experimental modems and pertinentsoftware across some or all layers of theprotocol stack

• Efforts that involve primarily softwaredevelopment for the network and applica-tion layers of the protocol stack of UANsthat use commercially available modems

The majority of existing proposals offer onlydata logging capabilities for offline processingand do not report real-time reconfigurability andonline performance evaluation through experi-mental field studies.

The main objective of high-layer software pro-posals in underwater networks is the develop-ment of a framework that facilitates integrationof simulation, emulation, and testing policies andcan interface with existing commercial or experi-mental underwater modems. The work in [2]reviews and compares two software frameworks,SUNSET and DESERT, implemented using theopen source network simulators ns-2 and ns-2miracle. SUNSET has been used as the standardplug-in in the SUNRISE [3] project. SUNRISE isa federation of testing infrastructures that isdesigned to represent different marine environ-ments and support different applications. Ocean-TUNE [4] presents a similar testbed suite thatprovides flexibility in system and network config-urations. Aqua-Net [5] is a software frameworkthat uses WHOI micro-modems and TeledyneBenthos modems, and offers a layered architec-ture that enables cross-layer optimization. Thework in [6] proposes an agent-based (service-ori-ented) architecture on a Java Virtual Machine(JVM), therefore allowing portability from simu-lation to real-time deployment. A more recentwork [7] attempts to depart from OSI-layeredimplementations and follow a modular cross-layer software-defined architecture.

Lower-layer/hardware proposals in underwa-ter networks concentrate on hardware develop-ment of underwater modems and explore theirinterface with hybrid software architectures thatoffer reconfigurability at the PHY and MAC lay-ers. In particular, the work in [8] studies amodem architecture that is based on the open-source software tools GNU Radio, TinyOS, andTOSSIM, and the USRP hardware. The work in

There is a need for

UAN devices that

can i) ease deploy-

ment and testing of

new protocol

designs, ii) resolve

the interoperability

problem in heteroge-

neous UANs, and iii)

intelligently decide

and adapt their com-

munication parame-

ters in real time.


IEEE Communications Magazine • November 201566

[8] builds an underwater sensor network moteand parallels pertinent developments in the soft-ware radio and sensor network literature. Simi-larly, [9–11] provide early modem prototypes thatare based on either FPGA/DSP or FPGA-onlytechnologies and develop in-house software forcontrolling the PHY and MAC layer parameters.

PROPOSED SDAM PROTOTYPEIn this section, we present our developmentstoward the design and implementation of a vari-able-rate fully reconfigurable SDAM that is builtfrom a collection of COTS components. Wedemonstrate and evaluate, in real-time underwa-ter field experiments, the software-defined capa-bilities of the proposed modem/design. Weprovide a complete overview (from hardware tosoftware) of the SDAM implementation, as wellas our design considerations for the PHY, MAC,and network layers of the protocol stack.

HARDWARE SETUPThe SDAM is at its core an SDR connected withwideband acoustic transducers through poweramplifiers/preamplifiers. The SDAM exploits theunique capabilities and features of an SDR tofulfill the need for flexible, easily reconfigurableUAN devices [12].

The overall hardware architecture of the pro-posed SDAM prototype is illustrated in Fig. 1,while Fig. 2 depicts the prototype itself. The pro-posed SDAM is based on a USRP N210, which isa commercially available FPGA-based, SDR plat-form. We chose to work with LFTX and LFRXdaughterboards (DC –30 MHz), which enablethe development of a half-duplex transceiveroperating in the frequency range of the selectedomnidirectional acoustic transducer, TeledyneRESON TC4013, from 1 Hz to 170 kHz. Toenhance the communication range of our SDAM,we used a linear wideband power amplifier (PA),Benthowave BII–5002, and a voltage preamplifier(PreA), Teledyne RESON VP2000. We alsoincorporated an electronic switch, Mini-CircuitsZX80–DR230+ to enable the operation of a sin-

gle acoustic transducer as transmitter and receiv-er in a time-division duplex fashion. Basebandsignal processing algorithms and protocols aremainly implemented in the host PC, which isconnected to the USRP N210 through GigabitEthernet (GigE). Each prototype costs approxi-mately $6000 (commercially available alternativesare almost 2× more expensive).

PHYSICAL LAYERWe consider the design and implementation oftwo different signaling technologies for differentapplications based on zero-padded orthogonalfrequency-division multiplexing (ZP-OFDM)and direct sequence spread-spectrum (DS-SS).Both technologies aim to maximize channel uti-lization and spectral efficiency in the challengingunderwater environment by providing onlineadaptation at both PHY and MAC layers.

ZP-OFDM: We adopt a ZP-OFDM with asuperimposed convolutional error correctioncoding scheme. In particular, we use a K-subcar-rier ZP-OFDM where zero-padding works as azero-power guard interval alternative to the con-ventional cyclic prefix. Before subcarrier alloca-tion, data symbols are modulated with eitherbinary phase shift keying (BPSK) or quadraturePSK (QPSK), while KP = K/4 are used as pilot,KD as data, and KN as null subcarriers. A guardinterval is added between each OFDM symbolto avoid interference among different OFDMsymbols, and a pseudorandom noise (PN)sequence is transmitted at the beginning of eachOFDM packet for frame detection and coarsesynchronization purposes at the receiver. At thereceiver side, a low-pass filter (LPF) is first usedto reject out-of-band noise and interference.Null subcarriers are used for Doppler scale esti-mation and compensation, while pilot subcarri-ers are used for channel estimation and finesymbol synchronization in the frequency domain.

DS-SS: In this scheme, BPSK data symbolsare modulated by a random binary code-wave-form of length L and duration 1/(B – A); thus,transmitted waveforms occupy the whole avail-able bandwidth (A Hz, B Hz) of the device, pro-viding efficient utilization of the underwateracoustic spectrum resources. To avoid inter-sym-bol interference we ensure that the productbetween code length L and waveform duration islarger than the multipath spread; hence, wechoose large code lengths, which leverage thecapability of the selected technique for multipleaccess. More specifically, we adopt two differenttransmitter designs, one where a long PNsequence is used for frame detection, synchro-nization, and channel estimation at the receiverside, and a second where an all-1 sequence ofunmodulated bits is used to achieve frame detec-tion and coarse synchronization at the receiver[13]. For the second transmitter setup, we pro-pose a blind receiver design that integrates andexecutes synchronization, channel estimation,and demodulation in a combined fashion with alow-complexity RAKE-type receiver [13]. Theproposed receiver structure offers significantcomputational efficiency and resistance to multi-ple access interference, therefore sparking inter-est in future applications on heavily utilizedspectrum resources.

Figure 1. Hardware architecture of the proposed SDAM prototype.

USRP N-210

FPGA-Xilinx Spartan 3A-DSP

UHDnetwork driver

command & control

data streaming

Decim

InterCIC

DDC

DUC DAC

ADC LFRX

LFTX

PreA

PA

SWITCH

Acoustic transducer

Gigabit Ethernet

GNU Radio/MATLAB

HOST PC



USER-DEFINED DECISION ALGORITHMS

Herein, we define the decision algorithms foradaptation in both the ZP-OFDM and DS-SSphysical layer schemes. However, we first need todefine a reliable feedback communication methodto support real-time forward link adaptation, aspotential unsuccessful feedback delivery mayresult in failure of the forward link as well. Tothat end, we propose and implement a feedbackmethod based on binary chirp spread-spectrum(B-CSS) modulation, which is known to beresilient against the severe multipath and Dopplereffects that characterize the UW-A channel. Inaddition, B-CSS provides a robust low data ratecommunication scheme with a low-complexityreceiver design that very well fits the reliabilityand robustness requirements of an underwaterfeedback channel. More specifically, we leveragethe quasi-orthogonality of up and down chirps byencoding a “1” bit with an up chirp and a “0” bitwith a down chirp. Up and down chirp signals aregenerated by proper selection of the time-varyinginstantaneous frequency that ranges from f0 to f1,and the period T of a chirp signal. Therefore, ifthe chirp frequency variation rate defined as m =(f1 – f0)/T is positive, we generate an up chirp;otherwise, if m < 0, we generate a down chirp.

Decision algorithms provide the SDAM withthe capability to either adapt/change specificcommunication parameters (of a pre-selectedcommunication technology) or switch betweendifferent communication technologies such asZP-OFDM and DS-SS. Decisions are made bythe receiver node based on user-defined algo-rithms, and are then communicated to the trans-mitter through the wireless feedback link. Weconsider three decision algorithms that best illus-trate the adaptation capabilities of the proposedSDAM under preset performance constraints.

First, we consider adaptation of the modula-tion and the error correction coding rate in aZP-OFDM link to solve a rate maximizationproblem under preset BER reliability constraints.We assume that the number of data subcarriers,as well as the symbol and guard interval durationin ZP-OFDM remain fixed, and data rate is afunction of the modulation order and error cor-rection coding rate. The receiver estimates thesignal-to-interference-plus-noise ratio (SINR)per received packet, and jointly decides on thecombination of a modulation (BPSK or QPSK)and error correction coding (1/2-rate convolu-tional code or uncoded) scheme to satisfy a pre-defined BER threshold constraint.

Second, we provide the SDAM with the capa-bility to switch between ZP-OFDM and DS-SScommunication technologies. This is achieved bydesigning a multiplexer-based adaptation mecha-nism that is activated upon successful decodingof an incoming packet. The SDAM is prepro-grammed with all the primitive modules requiredfor the implementation of both ZP-OFDM andDS-SS physical layer schemes. The enabling sig-nal of the multiplexer is controlled by a Booleanvariable that enables DS-SS transmission/recep-tion upon successful reception of a ZP-OFDMpacket and vice versa.

Finally, we consider real-time code-waveformadaptation in a DS-SS link to maximize the pre-

detection SINR at the destination receiver ofinterest in the presence of multiple access inter-ference. The receiver first senses the environmentand calculates a disturbance (noise plus interfer-ence) autocorrelation matrix during the time thatthe user of interest is silent. Then the receiverestimates the pre-detection SINR for each incom-ing packet and looks for the best channel wave-form of length L that is the solution to an SINRmaximization problem. The new channel wave-form is communicated to the transmitting sourceof interest through a chirp-based feedback link.Both the transmitter and receiver need to be syn-chronized. We implement two parallel processingreceiver chains, one working with the updatedand another with the preset channel waveform.

MEDIUM ACCESS CONTROLWe implement a simple MAC protocol that cansupport user-defined decision algorithms. Specifi-cally, at the time an SDAM node has data avail-able for transmission, it directly accesses thechannel, transmits a data packet, and switches intoreceive mode, waiting for a feedback message fromthe destination node. A timeout time is also set forretransmission. In the absence of data, all nodesperform idle listening. Even though the currentimplementation includes one simple MAC proto-col design, several others can be implemented byexploiting primitive built-in functionalities such astimer operations, idle listening, and retransmission.

NETWORK LAYERThe proposed SDAM prototype can supportIPv4 and IPv6 protocols through an adaptationlayer [14]. The adaptation layer offers a set offunctionalities, i.e., IP header compression, IPpacket fragmentation that enable the interfacingof the traditional IP network layer with the MAClayer of the prototype. Specifically, the tradition-al IPv4 and IPv6 headers are optimized forunderwater acoustic communications to minimizethe network delay and energy consumption.

SOFTWARE DEVELOPMENTSWe use an open source software framework calledGNU Radio to drive adaptive baseband signalprocessing. GNU Radio offers a plethora of digi-

Figure 2. The proposed SDAM prototype.



tal signal processing blocks that are implementedin C++, and are usually wrapped into Pythonclasses. Blocks are either instantiated by Pythonscripts or used as building primitives of a commu-nication flowgraph. GNU Radio provides userswith the ability to design custom-logic signal pro-cessing blocks, simulate them offline, and finallyembed them into existing flowgraph designs. Inthis context, we design and implement custom sig-nal processing blocks and build ZP-OFDM andDS-SS flowgraphs, while at the same time exploitthe message passing capabilities of GNU Radioto support adaptive transmitter and receiver phys-ical layer functionalities. Message passing allowsthe exchange of control messages between differ-ent blocks that are located either downstream orupstream in the GNU Radio flowgraph. There-fore, we achieve real-time adaptation by leverag-ing both GNU Radio’s asynchronous messagingfeatures and the properties of a chirp-based feed-back link. Decoding of the feedback messagesrelies on two correlation filters equipped with anup chirp and a down chirp, respectively. Wedecide on the feedback message bits by compar-ing the outputs of the respective correlation filtersat the receiver, and we update the PHY layerparameters in the forward link flowgraph accord-ingly. For the needs of communication technologyadaptation between ZP-OFDM and DS-SS, wealso implement a software multiplexer block.

EXPERIMENTSIn this section, we present experimental resultsfrom three different testbed scenarios and evalu-ate the proposed SDAM prototype in bothindoor and outdoor environments.

PHYSICAL LAYER ONLINE ADAPTATIONA series of experiments took place in LakeLaSalle at the State University of New York atBuffalo. Lake LaSalle has a depth of approxi-mately 7 ft. We deployed two SDAM prototypes322 ft apart from each other as illustrated in Fig.3. We used ZP-OFDM signals that occupy abandwidth of B = 24 kHz at a carrier frequencyfc = 100 kHz. We defined K = 1024 subcarriers

for each OFDM symbol, where each subcarrieris either mapped with BPSK or QPSK modula-tion and is either coded (rate 1/2 convolutionalerror correction codes) or not coded. We alsoincorporated a guard time of Tg = 15 msbetween each OFDM symbol.

The objective of this set of experiments is todemonstrate the real-time adaptation capabili-ties of our SDAM prototype at the physical layerin an outdoor lake environment. Figure 4 depictsexperimental data rate and BER evaluationresults of both an adaptive modulation/codingrate and a fixed (non-adaptive) scheme for 21OFDM packet transmissions. The modulationscheme and coding rate of ZP-OFDM changeaccording to a user-defined decision algorithmthat aims to maximize data rate under presetBER threshold constraints. In Fig. 4 the BERthreshold is empirically set to 10–3, while SINRvaries between 10 and 20 dB.

In both adaptive and fixed (non-adaptive)schemes, the SDAM, N1, starts transmitting atthe highest possible data rate, which requiresuncoded QPSK modulation. We observe thatsince the non-adaptive scheme does not providethe SDAM with modulation/coding adaptationcapabilities, data rate will remain constant. How-ever, at the time that the SINR profile changes(6th packet), the preset BER threshold is notsatisfied. On the other hand, in the adaptivescheme, the SDAM adapts the modulation andcoding rate in real time as soon as the SINRdecreases. Specifically, when the estimated SINRis lower than 10 dB, the SDAM incorporates1/2-rate error correction coding and changes themodulation scheme from QPSK to BPSK. As aresult, the data rate is adjusted to a lower valueto satisfy the predefined BER constraints. Subse-quently, as soon as the SINR increases (frompacket 11 to 17), the SDAM changes into uncod-ed BPSK transmission to compensate for thedata rate loss. Finally, the SDAM switches backto uncoded QPSK when the SINR reaches theinitial level of 20 dB.

COGNITIVE CHANNELIZATIONThe following set of experiments considers thedeployment of three SDAMs in a water testtank. Figure 5 depicts three SDAM nodes, whichoperate in the same frequency band, 91–99 kHz,and use random binary waveforms of length L =63. N1, and N2 act as transmitters, and N3 as areceiver, while feedback messages regarding thebest channel waveform are exchanged betweenN1 and N3. Both transmitters use square-rootraised cosine pulses of duration Td = 0.16384ms, and roll-off factor a = 0.35. The number ofchannel paths is set to N =20 and we measureda multipath spread of 31 waveform bits or 5.1ms. Both N1 and N2 use the same transmissionpower and are positioned 5.5 ft apart from N3.The distance between N1 and N2 is set at 1.5 ft.

Figure 6 illustrates SINR experimental stud-ies for two different setups. The first setup con-siders a single transmitter node, N1, thatoperates in the absence of N2. The second setupconsiders two nodes, N1 and N2, that transmitsimultaneously at the same frequency band,while a third node, N3, uses a control channel ata different frequency band to periodically com-

Figure 3. Underwater transducers are held bythe red buoys and are deployed 322 ft apart.

Teledyne Reson TC4013 - N2

Teledyne Reson TC4013 - N1

GNU Radio offers a

plethora of digital

signal processing

blocks which are

implemented in

C++, and are

usually wrapped into

Python classes.

Blocks are either

instantiated by

Python scripts,

or used as building

primitives of a

communication

flowgraph.



municate to N1 a new channel waveform. Foreach setup we consider two different scenariosreferred to as cognitive and fixed channelization,respectively. Cognitive channelization assumesthat the transmitter node N1 adapts its channelwaveform based on the interference profile ofN2 and the additive noise. In the absence of N2,waveform adaptation is based only on the noiseprofile. For the fixed channelization scenario, N1does not change the preset waveform. Weobserve that for both setups, cognitive channel-ization significantly outperforms fixed channel-ization in terms of receiver predetection SINR.

The experimental studies in this section showthat underwater cognitive channelization is ben-eficial in both the presence and absence of inter-ference from other signals/users due to the factthat the SDAM efficiently adapts to channelconditions [13].

COMMUNICATION TECHNOLOGY ADAPTATION(FUTURE HETEROGENEOUS UW NETWORKS)

Given the primitives of both ZP-OFDM and DS-SS, we also tested the capability of the proposedSDAM prototype to adapt/choose among differ-ent communication technologies. We designedan experimental scenario that aims at seamlessswitching between OFDM packet (BPSK modu-lation, no error correction coding, K = 1024 sub-carriers in a bandwidth of 24 kHz) and DS-SSpacket (BPSK modulation, no error correctioncoding, and waveform length L = 31 in a band-width of 6 kHz) transmissions at carrier frequen-cy fc = 100 kHz, and successfully demonstratedthe dual mode capability of the prototype.

SOFTWARE-DEFINED UANS:OPEN RESEARCH CHALLENGES

Reconfigurability is an essential feature of next-generation underwater networks in early stagesof study, especially in the context of real-time

joint adaptation across all layers of the protocolstack. Efforts toward reconfigurable underwaternetworks can certainly benefit from recent andparallel developments in the SDR literature.The adoption of SDR frameworks in underwatercommunications must carefully consider thechallenges imposed by both the available tech-nology and the underwater environment. Below,we present an overview of the core challengesrelated to software-defined UAN implementa-tions.

HARDWARE CHALLENGESExisting commercial or military SDR platformsprovide broadband frequency support (DC — 6GHz) with either fixed or configurable front-ends. However, the majority of the designs areoptimized for RF terrestrial communications,and cannot address the challenges of the under-water environment (e.g., acoustic frequencies onthe order of kilohertz). In addition, SDR front-ends that can be used directly for underwatercommunications utilizing the LFTX daughter-card on a USRP platform, which results in oper-ating frequencies in the range of DC to 30 MHz,

Figure 5. Cognitive channelization water testtank setup.

Figure 4. Comparison of adaptive with fixed (non-adaptive) scheme in terms of data rate, and BER fordifferent SINR profiles.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 210

10

20

30

Dat

a ra

te (

kb/s

)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 210

10

20

30

SIN

R (d

B)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 210

0.02

0.04

Packet index

Bit

erro

r ra

te

FixedAdaptive

FixedAdaptive

Fixed, QPSK, uncodedAdaptive, QPSK, uncodedAdaptive, BPSK, 1/2 codedAdaptive BPSK, uncoded

Efforts towards

reconfigurable

underwater net-

works can certainly

benefit from recent

and parallel develop-

ments in the SDR lit-

erature. The

adoption of SDR

frameworks in

underwater commu-

nications must care-

fully consider the

challenges imposed

both by the available

technology and the

underwater

environment.

DEMIRORS_LAYOUT.qxp_Author Layout 10/30/15 2:54 PM 9


do not include a power amplifier; thus, underwa-ter communication is limited in terms ofrange/distance. Furthermore, commercial off-the-shelf (COTS) acoustic transducers are expen-sive, (their cost is tens of thousands of dollars)due to a proprietary and long process of charac-terization and analog circuitry optimization.COTS transducers also exhibit application-spe-cific characteristics and low-volume production,and inherently have limited bandwidth capabili-ties. As a result, experimental prototypingrequires careful review of the available COTShardware components and specifications, andmay also require the design of a custom circuitfront-end.

SOFTWARE PORTABILITYOne of the major design challenges of SDR orsoftware-defined underwater systems is thedevelopment of platform-independent software.Effective proposals need to be able to demon-strate that pertinent software can easily berebuilt for different SDR platforms with mini-mum effort/cost. Software interoperability isespecially desirable in military systems wherestandardization is required for both the inter-connection between application components,and between application components and sys-tem devices. To date, we do not have a com-plete structured (abstract) methodology/architecture available to implement reconfigura-tions across all layers of the protocol stack. Suchan architecture could benefit future underwatercommunications by supporting, for example,cognitive functionalities that may lead to highefficiency in spectrum utilization. For instance,software components for static underwaterwaveform processing could be real-timeswapped with cognitive components for adaptiveprocessing to mitigate adverse external condi-tions (e.g., noise and multiple-access interfer-ence).

COMPUTATIONAL CAPACITY

Future wideband high-bit-rate underwater com-munications require significant computationalresources. It is well understood that data ratecapabilities are coupled with the front-end designof the software-defined underwater networknodes (e.g., acoustic transducers and poweramplifiers) and may vary according to the exter-nal environment as well as the role of the under-water software-defined unit in the network (e.g.,surface station, AUV, etc.), which poses differentweight, cost, and power consumption constraints.Similar to wired computer networks, software-defined underwater acoustic networks may bene-fit from a separation between data processingand control functionalities. As a result, futuresoftware-defined modem architectures could beable, for example, to distribute data processingbetween hardware and software and exploit par-allelism for computation-intensive data process-ing components (e.g., finite impulse responsefilters, fast Fourier transform modules).

ENERGY EFFICIENCYEnergy consumption of submerged units is a sig-nificant design challenge for future reconfig-urable underwater modems. It is evident thatalgorithmic implementation on architecturesbased on application-specific integrated circuits(ASICs) or DSP-on-chip trade reconfigurabilityfor computational and power efficiency. Recent-ly, a new class of technologies such as pro-grammable logic and IP (FPGAs), in-line vectorinstructions (ARM NEON), and vector execu-tion units (modern GPUs) are breaking theboundaries of device functionalities and aim atcreating a multicore blend of functions, as seen,for example, in several new products includingTI KeyStone, Xilinx Zynq, and NVIDIA TegraK1. Multicore architectures require softwareabstractions to control reconfiguration of tai-lored functional units or general processing com-ponents. For example, the work in [15]implements matching pursuit algorithms forunderwater channel estimation in FPGAs(reconfigurable hardware platforms) and pro-vides 210× and 52× reduction in energy con-sumption over the microcontroller and DSPimplementations, respectively.

REAL-TIME RECONFIGURABLE PROCESSINGWhile considerable research work for the next-gen-eration processors aim at increasing cycle clocks,mutli-core architecture proposals (i.e., single-instruction multiple-data and multiple-instructionmultiple-data) allow a higher average number ofinstructions to be processed per clock cycle, andoffer high performance and reconfiguration byexploiting parallelism. However, even thoughmulti-core DSPs offer optimized features for digi-tal signal processing, we still need to study thetrade-off between high performance and low powerconsumption in software-defined underwateracoustic networks. More specifically, reconfigura-tion timing constraints can be significantly relaxeddue to larger channel coherence times in underwa-ter communication (on the order of seconds) com-pared to wireless LAN communications (on theorder of milliseconds). Relaxed timing constraints

Figure 6. Cognitive channelization is compared to fixed channelization interms of predetection SINR. The depicted SINR gains consider two dif-ferent setups: a blind receiver configuration in a multi-user case and in asingle-user (N1–N3) case.

0 100 200 300 400 500 600 700 800 900 10005

10

15

20

25

30

Time (s)

Pre-

dete

ctio

n SI

NR

(dB)

Fixed channelizationCognitive channelizationFixed channelization − single userCognitive channelization − single user

DEMIRORS_LAYOUT.qxp_Author Layout 10/30/15 2:54 PM 0


imply low power consumption, which is beneficialto battery operated underwater deployments.

CONCLUSIONSWe have discussed SDR technology principles forUANs that need reconfigurable and intelligent net-work devices. Accordingly, we have proposed anSDAM prototype, based on COTS components,and have experimentally evaluated the proposedprototype in both indoor (water test tank) and out-door (lake) environments. Then we have demon-strated, under three experimental scenarios, thereal-time reconfigurable capabilities of the pro-posed SDAM prototype and highlighted its favor-able characteristics toward spectrally efficientcognitive underwater networks and high data rateunderwater acoustic links. Finally, we havereviewed open research implementation challengesfor future next-generation software-defined UANsand presented an overview of existing SDR-basedarchitectures for UW-A communications.

ACKNOWLEDGMENTSThis work was supported in part by the NationalScience Foundation under grants CNS-1422874and CNS-1126357.

REFERENCES[1] B. Tomasi, “JANUS: The Genesis, Propagation and Use

of an Underwater Standard,” Euro. Conf. UnderwaterAcoustics, Istanbul, Turkey, July 2010.

[2] R. Petroccia and D. Spaccini, “Comparing the Sunsetand Desert Frameworks for In Field Experiments inUnderwater Acoustic Networks,” OCEANS — Bergen,2013 MTS/IEEE, June 2013, pp. 1–10.

[3] C. Petrioli et al., “The Sunrise Gate: Accessing the Sun-rise Federation of Facilities to Test Solutions for theInternet of Underwater Things,” Underwater Commun.and Networking 2014, Sept. 2014, pp. 1–4.

[4] J.-H. Cui et al., “Ocean-TUNE: A Community OceanTestbed for Underwater Wireless Networks,” ACM Int’l.Conf. Underwater Networks and Systems, Los Angeles,CA, Nov. 2012, pp. 1–2.

[5] Z. Peng et al., “Aqua-net: An Underwater Sensor Net-work Architecture: Design, Implementation, and InitialTesting,” OCEANS 2009, MTS/IEEE Biloxi — MarineTechnology for Our Future: Global and Local Chal-lenges, Oct. 2009, pp. 1–8.

[6] M. Chitre, R. Bhatnagar, and W.-S. Soh, “Unetstack: AnAgent-Based Software Stack and Simulator for Under-water Networks,” Oceans — St. John’s 2014, Sept2014, pp. 1–10.

[7] J. Potter et al., “Software Defined Open ArchitectureModem Development at CMRE,” Underwater Commun.and Networking 2014, Sept 2014, pp. 1–4.

[8] D. Torres et al., “Software-Defined Underwater AcousticNetworking Platform,” Proc. 4th ACM Int’l. Wksp.UnderWater Networks, ser. WUWNet ‘09, Nov 2009,pp. 7:1–7:8.

[9] E. M. Sözer and M. Stojanovic, “Reconfigurable Acous-tic Modem for Underwater Sensor Networks,” Proc. 1stACM Int’l. Wksp. Underwater Networks, ser. WUWNet‘06, Sept 2006, pp. 101–04.

[10] N. Nowsheen, C. Benson, and M. Frater, “A High Data-Rate, Software-Defined Underwater Acoustic Modem,”OCEANS 2010, Sept 2010, pp. 1–5.

[11] M. Chitre, I. Topor, and T. Koay, “The UNET-2 Modem —An Extensible Tool for Underwater Networking Research,”OCEANS, 2012 — Yeosu, May 2012, pp. 1–7.

[12] E. Demirors et al., “Design of a Software-DefinedUnderwater Acoustic Modem with Real-Time PhysicalLayer Adaptation Capabilities,” Proc. Int’l. Conf. Under-water Networks & Systems, ser. WUWNET ‘14, 2014,pp. 25:1–25:8.

[13] G. Sklivanitis et al., “Receiver Configuration andTestbed Development for Underwater Cognitive Chan-nelization,” 2014 48th Asilomar Conf. Signals, Systemsand Computers, Nov 2014, pp. 1594–98.

[14] Y. Sun and T. Melodia, “The Internet Underwater: An IP-Compatible Protocol Stack for Commercial UnderseaModems,” Proc. 8th ACM Int’l. Conf. Underwater Networksand Systems, ser. WUWNet ‘13, 2013, pp. 37:1–37:8.

[15] B. Benson et al., “Energy Benefits of ReconfigurableHardware for Use in Underwater Snesor Nets,” IEEEInt’l. Symp. Parallel Distrib. Processing 2009, May 2009,pp. 1–7.

BIOGRAPHIESEMRECAN DEMIRORS [S’11] ([email protected]) receivedhis B.Sc. degree in electrical and electronics engineeringfrom the Bilkent University, Ankara, Turkey, in 2009 and hisM.Sc. degree in electrical engineering from the State Uni-versity of New York at Buffalo, New York, in 2013. He iscurrently working toward his Ph.D. degree in electricalengineering at the Department of Electrical and ComputerEngineering at Northeastern University, Boston, Mas-sachusetts. His research interests include software-definedwireless communications and networks, cognitive radionetworks, and underwater acoustic communications andnetworks. In 2014, he was the winner of the Nutaq Soft-ware-Defined Radio Academic US National Contest.

GEORGE SKLIVANITIS [S’11] ([email protected]) receivedhis Diploma in electronic and computer engineering fromthe Technical University of Crete, Chania, Greece, in 2010.He is currently working toward his Ph.D. degree in electri-cal engineering at the State University of New York at Buf-falo. His research interests span the areas of signalprocessing, software-defined wireless communications andnetworking, cognitive radio, and underwater acoustic com-munications. In 2014 he was the winner of the Nutaq Soft-ware-Defined Radio Academic US National Contest, and in2015 he received the Graduate Student Excellence inTeaching award from University at Buffalo.

TOMMASO MELODIA [M’07] ([email protected]) received hisPh.D. degree in electrical and computer engineering from theGeorgia Institute of Technology, Atlanta, in 2007. He is anassociate professor with the Department of Electrical andComputer Engineering, Northeastern University, Boston, Mas-sachusetts. He serves on the Editorial Boards of IEEE Transac-tions on Mobile Computing, IEEE Transactions on WirelessCommunications, IEEE Transactions on Multimedia, and Com-puter Networks. His research interests are in underwater sen-sor networks, cognitive and software-defined networks, andintra-body medical networks of implantable devices.

STELLA N. BATALAMA [S’91, M’94] ([email protected])received her Diploma degree in computer engineering andscience from the University of Patras, Greece, in 1989 andher Ph.D. degree in electrical engineering from the Universityof Virginia, Charlottesville, in 1994. In 1995 she joined theDepartment of Electrical Engineering, State University ofNew York at Buffalo, where she is presently a professor.From 2009 to 2011, she served as the associate dean forresearch of the School of Engineering and Applied Sciences,and since 2010 she has served as chair of the Electrical Engi-neering Department. During the summers of 1997–2002 shewas visiting faculty at the U.S. Air Force Research Laboratory(AFRL), Rome, New York. From August 2003 to July 2004she served as acting director of the AFRL Center for Integrat-ed Transmission and Exploitation (CITE). Her research inter-ests include small-sample-support adaptive filtering andreceiver design, cooperative communications, cognitive net-works, underwater communications, covert communications,steganography, compressive sampling, adaptive multiuserdetection, robust spread-spectrum communications, andsupervised and unsupervised optimization. She was an Asso-ciate Editor for IEEE Communications Letters (2000–2005)and IEEE Transactions on Communications (2002–2008).

DIMITRIS A. PADOS [M’95, SM’15] ([email protected])received his Diploma degree in computer science and engi-neering from the University of Patras and his Ph.D. degreein electrical engineering from the University of Virginia.Since August 1997, he has been with the Department ofElectrical Engineering, State University of New York at Buf-falo, where he currently holds the title of Clifford C. FurnasProfessor of Electrical Engineering. His research interestsare in the general areas of communication theory and sys-tems, and adaptive signal processing with applications tointerference channels and signal waveform design, securewireless communications, and cognitive acoustic-to-RFmodems and networks.

Reconfiguration tim-

ing constraints can

be significantly

relaxed due to larger

channel coherence

times in underwater

communication (on

the order of seconds)

compared to wireless

LAN communications

(on the order of mil-

liseconds). Relaxed

timing constraints

imply low power

consumption, which

is beneficial to

battery operated

underwater

deployments.


software-defined underwater acoustic networks: …gsklivan/underwatermag15.pdf · software-defined...

Documents