Underwater Acoustic Sensor Networks: Research Challenges

260I.F. Akyildiz et al. / Ad Hoc Networks 3 (2005) 257–279and ongoing research projects related to underwater communications and explorations.The remainder of this paper is organized asfollows. In Sections 2 and 3 we introduce the communication architecture and design challenges,respectively, of underwater acoustic sensornetworks. In Section 4, we investigate the underwater acoustic communication channel andsummarize the associated physical layer challengesfor underwater networking. In Sections 5–9 we discuss physical, data link, network, transport andapplication layer issues in underwater sensor networks, respectively. In Section 10 we describesome experimental implementations of underwatersensor networks while in Section 11 we draw themain conclusions.2. Underwater acoustic sensor networks:communication architectureIn this section, we describe the communicationarchitecture of underwater acoustic sensornetworks. In particular, we introduce referencearchitectures for two-dimensional and threedimensional underwater networks, and presentseveral types of autonomous underwater vehicles(AUVs) which can enhance the capabilities ofunderwater sensor networks.The network topology is in general a crucialfactor in determining the energy consumption, thecapacity and the reliability of a network. Hence,the network topology should be carefully engineered and post-deployment topology optimizationshould be performed, when possible.Underwater monitoring missions can be extremely expensive due to the high cost of underwaterdevices. Hence, it is important that the deployednetwork be highly reliable, so as to avoid failureof monitoring missions due to failure of single ormultiple devices. For example, it is crucial to avoiddesigning the network topology with single pointsof failure that could compromise the overall functioning of the network.The network capacity is also influenced by thenetwork topology. Since the capacity of the underwater channel is severely limited, as will be discussed in Section 4, it is very important toorganize the network topology such a way thatno communication bottleneck is introduced.The communication architectures introducedhere are used as a basis for discussion of the challenges associated with underwater acoustic sensornetworks. The underwater sensor network topology is an open research issue in itself that needsfurther analytical and simulative investigationfrom the research community. In the remainderof this section, we discuss the followingarchitectures: Static two-dimensional UW-ASNs for ocean bottom monitoring. These are constituted by sensornodes that are anchored to the bottom of theocean, as discussed in Section 2.1. Typicalapplications may be environmental monitoring,or monitoring of underwater plates in tectonics[21]. Static three-dimensional UW-ASNs for oceancolumn monitoring. These include networks ofsensors whose depth can be controlled by meansof techniques discussed in Section 2.2, and maybe used for surveillance applications or monitoring of ocean phenomena (ocean bio–geochemical processes, water streams, pollution). Three-dimensional networks of autonomousunderwater vehicles (AUVs). These networksinclude fixed portions composed of anchoredsensors and mobile portions constituted byautonomous vehicles, as detailed in Section 2.3.2.1. Two-dimensional underwater sensor networksA reference architecture for two-dimensionalunderwater networks is shown in Fig. 1. A groupof sensor nodes are anchored to the bottom ofthe ocean with deep ocean anchors. Underwatersensor nodes are interconnected to one or moreunderwater sinks (uw-sinks) by means of wirelessacoustic links. Uw-sinks, as shown in Fig. 1, arenetwork devices in charge of relaying data fromthe ocean bottom network to a surface station.To achieve this objective, uw-sinks are equippedwith two acoustic transceivers, namely a verticaland a horizontal transceiver. The horizontal transceiver is used by the uw-sink to communicate with

I.F. Akyildiz et al. / Ad Hoc Networks 3 (2005) 257–279261Fig. 1. Architecture for 2D underwater sensor networks.the sensor nodes in order to: (i) send commandsand configuration data to the sensors (uw-sink tosensors); (ii) collect monitored data (sensors touw-sink). The vertical link is used by the uw-sinksto relay data to a surface station. In deep waterapplications, vertical transceivers must be longrange transceivers as the ocean can be as deep as10 km. The surface station is equipped with anacoustic transceiver that is able to handle multipleparallel communications with the deployed uwsinks. It is also endowed with a long range RFand/or satellite transmitter to communicate withthe onshore sink (os-sink) and/or to a surface sink(s-sink).Sensors can be connected to uw-sinks via directlinks or through multi-hop paths. In the formercase, each sensor directly sends the gathered datato the selected uw-sink. However, in UW-ASNs,the power necessary to transmit may decay withpowers greater than two of the distance [44], andthe uw-sink may be far from the sensor node.Consequently, although direct link connection isthe simplest way to network sensors, it may notbe the most energy efficient solution. Furthermore, direct links are very likely to reduce the net-work throughput because of increased acousticinterference due to high transmission power. Incase of multi-hop paths, as in terrestrial sensornetworks [10], the data produced by a source sensor is relayed by intermediate sensors until itreaches the uw-sink. This may result in energysavings and increased network capacity, but increases the complexity of the routing functionality. In fact, every network device usually takespart in a collaborative process whose objective isto diffuse topology information such that efficientand loop free routing decisions can be made ateach intermediate node. This process involves signaling and computation. Since energy and capacityare precious resources in underwater environments,as discussed above, in UW-ASNs the objective isto deliver event features by exploiting multi-hoppaths and minimizing the signaling overhead necessary to construct underwater paths at the sametime.2.2. Three-dimensional underwater sensor networksThree dimensional underwater networks areused to detect and observe phenomena that cannot

262I.F. Akyildiz et al. / Ad Hoc Networks 3 (2005) 257–279be adequately observed by means of ocean bottomsensor nodes, i.e., to perform cooperative samplingof the 3D ocean environment. In three-dimensional underwater networks, sensor nodes float atdifferent depths in order to observe a given phenomenon. One possible solution would be toattach each uw-sensor node to a surface buoy, bymeans of wires whose length can be regulated soas to adjust the depth of each sensor node [15].However, although this solution allows easy andquick deployment of the sensor network, multiplefloating buoys may obstruct ships navigating onthe surface, or they can be easily detected anddeactivated by enemies in military settings. Furthermore, floating buoys are vulnerable to weatherand tampering or pilfering.For these reasons, a different approach can beto anchor sensor devices to the bottom of theocean. In this architecture, depicted in Fig. 2, eachsensor is anchored to the ocean bottom andequipped with a floating buoy that can be inflatedby a pump. The buoy pushes the sensor towardsthe ocean surface. The depth of the sensor can thenbe regulated by adjusting the length of the wirethat connects the sensor to the anchor, by meansof an electronically controlled engine that resideson the sensor. A challenge to be addressed in suchan architecture is the effect of ocean currents onthe described mechanism to regulate the depth ofthe sensors.Many challenges arise with such an architecture, that need to be solved in order to enable3D monitoring, including: Sensing coverage. Sensors should collaboratively regulate their depth in order to achieve3D coverage of the ocean column, accordingto their sensing ranges. Hence, it must be possible to obtain sampling of the desired phenomenon at all depths. Communication coverage. Since in 3D underwater networks there may be no notion of uw-sink,sensors should be able to relay information tothe surface station via multi-hop paths. Thus,network devices should coordinate their depthsin such a way that the network topology isalways connected, i.e., at least one path fromevery sensor to the surface station always exists.Fig. 2. Architecture for 3D underwater sensor networks.

I.F. Akyildiz et al. / Ad Hoc Networks 3 (2005) 257–279Sensing and communication coverage in a 3Denvironment are rigorously investigated in [40].The diameter, minimum and maximum degree ofthe reachability graph that describes the networkare derived as a function of the communicationrange, while different degrees of coverage for the3D environment are characterized as a function ofthe sensing range. These techniques could beexploited to investigate the coverage issues inUW-ASNs.2.3. Sensor networks with autonomousunderwater vehiclesAUVs can function without tethers, cables, orremote control, and therefore they have a multitude of applications in oceanography, environmental monitoring, and underwater resourcestudy. Previous experimental work has shown thefeasibility of relatively inexpensive AUV submarines equipped with multiple underwater sensorsthat can reach any depth in the ocean [2]. Hence,they can be used to enhance the capabilities ofunderwater sensor networks in many ways. Theintegration and enhancement of fixed sensor networks with AUVs is an almost unexplored research area which requires new networkcoordination algorithms such as: Adaptive sampling. This includes control strategies to command the mobile vehicles to placeswhere their data will be most useful. Thisapproach is also known as adaptive samplingand has been proposed in pioneering monitoringmissions such as [4]. For example, the densityof sensor nodes can be adaptively increasedin a given area when a higher sampling rate isneeded for a given monitored phenomenon. Self-configuration. This includes control procedures to automatically detect connectivity holesdue to node failures or channel impairment andrequest the intervention of an AUV. Furthermore, AUVs can either be used for installationand maintenance of the sensor network infrastructure or to deploy new sensors. They canalso be used as temporary relay nodes to restoreconnectivity.263One of the design objectives of AUVs is tomake them rely on local intelligence and lessdependent on communications from online shores[25]. In general, control strategies are needed forautonomous coordination, obstacle avoidanceand steering strategies. Solar energy systems allowincreasing the lifetime of AUVs, i.e., it is not necessary to recover and recharge the vehicle on a daily basis. Hence, solar powered AUVs can acquirecontinuous information for periods of time of theorder of months [27].Several types of AUVs exist as experimentalplatforms for underwater experiments. Some ofthem resemble small-scale submarines (such asthe Odyssey-class AUVs [2] developed at MIT).Others are simpler devices that do not encompasssuch sophisticated capabilities. For example, drifters and gliders are oceanographic instruments often used in underwater explorations. Drifterunderwater vehicles drift with local current andhave the ability to move vertically through thewater column. They are used for taking measurements at preset depths [24]. Underwater gliders[18] are battery powered autonomous underwatervehicles that use hydraulic pumps to vary their volume by a few hundred cubic centimeters in orderto generate the buoyancy changes that power theirforward gliding. When they emerge on the surface,global positioning system (GPS) is used to locatethe vehicle. This information can be relayed tothe onshore station while operators can interactby sending control information to the gliders.Depth capabilities range from 200 m to 1500 mwhile operating lifetimes range from a few weeksto several months. These long durations are possible because gliders move very slowly, typically25 cm/s (0.5 knots). In [34], a control strategy forgroups of gliders to cooperatively move and reconfigure in response to a sensed distributed environment is presented. The proposed framework allowspreserving the symmetry of the group of gliders.The group is constrained to maintain a uniformdistribution as needed, but is free to spin and possibly wiggle with the current. In [20], results are reported on the application of the theory in [34] on afleet of autonomous underwater gliders during theexperiment on Monterey Bay in 2003 [4].

264I.F. Akyildiz et al. / Ad Hoc Networks 3 (2005) 257–2793. Underwater acoustic sensor networks: designchallengesIn this section, we describe the design challengesof underwater acoustic sensor networks. In particular, we itemize the main differences between terrestrial and underwater sensor networks, we detail keydesign issues and deployment challenges for underwater sensors, and we give motivations for a crosslayer design approach to improve the networkefficiency in the critical underwater environment.3.1. Differences with terrestrial sensor networksThe main differences between terrestrial andunderwater sensor networks are as follows: Cost. While terrestrial sensor nodes areexpected to become increasingly inexpensive,underwater sensors are expensive devices. Thisis especially due to the more complex underwater transceivers and to the hardware protectionneeded in the extreme underwater environment. Deployment. While terrestrial sensor networksare densely deployed, in underwater, thedeployment is deemed to be more sparse, dueto the cost involved and to the challenges associated to the deployment itself. Power. The power needed for acoustic underwater communications is higher than in terrestrial radio communications due to higherdistances and to more complex signal processing at the receivers to compensate for theimpairments of the channel. Memory. While terrestrial sensor nodes havevery limited storage capacity, uw-sensors mayneed to be able to do some data caching asthe underwater channel may be intermittent. Spatial correlation. While the readings from terrestrial sensors are often correlated, this is moreunlikely to happen in underwater networks dueto the higher distance among sensors.3.2. Underwater sensorsThe typical internal architecture of an underwater sensor is shown in Fig. 3. It consists of a mainFig. 3. Internal architecture of an underwater sensor node.controller/CPU which is interfaced with an oceanographic instrument or sensor through a sensorinterface circuitry. The controller receives datafrom the sensor and it can store it in the onboardmemory, process it, and send it to other networkdevices by controlling the acoustic modem. Theelectronics are usually mounted on a frame whichis protected by a PVC housing. Sometimes all sensor components are protected by bottom-mountedinstrument frames that are designed to permit azimuthally omnidirectional acoustic communications, and protect sensors and modems frompotential impact of trawling gear, especially inareas subjected to fishing activities. In [16], theprotecting frame is designed so as to deflect trawling gear on impact, by housing all componentsbeneath a low-profile pyramidal frame.Underwater sensors include sensors to measurethe quality of water and to study its characteristicssuch as temperature, density, salinity (interferometric and refractometric sensors), acidity, chemicals, conductivity, pH (magnetoelastic sensors),oxygen (Clark-type electrode), hydrogen, dissolvedmethane gas (METS), and turbidity. Disposablesensors exist that detect ricin, the highly poisonousprotein found in castor beans and thought to be apotential terrorism agent. DNA microarrays canbe used to monitor both abundance and activitylevel variations among natural microbial populations. Other existing underwater sensors includehydrothermal sulfide, silicate, voltammetric sensors

I.F. Akyildiz et al. / Ad Hoc Networks 3 (2005) 257–279for spectrophotometry, gold-amalgam electrodesensors for sediment measurements of metalions (ion-selective analysis), amperometric microsensors for H2S measurements for studies ofanoxygenic photosynthesis, sulfide oxidation, andsulfate reduction of sediments. In addition, force/torque sensors for underwater applications requiring simultaneous measurements of several forcesand moments have also been developed, as wellas quantum sensors to measure light radiationand sensors for measurements of harmful algalblooms.The challenges related to the deployment of lowcost, low scale underwater sensors, are listed asfollows: It is necessary to develop less expensive, robust‘‘nano-sensors’’, e.g., sensors based on nanotechnology, which involves development ofmaterials and systems at the atomic, molecular,or macromolecular levels in the dimensionrange of approximately 1–500 nm. It is necessary to devise periodical cleaningmechanisms against corrosion and fouling,which may impact the lifetime of underwaterdevices. For example, some sensors for pCO2,pH and nitrate measurement, and fluorometersand spectral radiometers, may be limited bybio-fouling, especially on a long time scale. There is a need for robust, stable sensors on ahigh range of temperatures since sensor driftof underwater devices may be a concern. To thisend, protocols for in situ calibration of sensorsto improve accuracy and precision of sampleddata must be developed. There is a need for new integrated sensors forsynoptic sampling of physical, chemical, andbiological parameters to improve the understanding of processes in marine systems.3.3. A Cross-layer protocol stackA protocol stack for uw-sensors should combine power awareness and management, and promote cooperation among the sensor nodes. Itshould consist of physical layer, data link layer,network layer, transport layer, and application265layer functionalities. The protocol stack shouldalso include a power management plane, a coordination plane, and a localization plane. The powermanagement plane is responsible for networkfunctionalities aimed at minimizing the energyconsumption (e.g., sleep modes, power control,etc.). The coordination plane is responsible forall functionalities that require coordination amongsensors (e.g., coordination of the sleep modes, dataaggregation, 3D topology optimization). Thelocalization plane is responsible for providingabsolute or relative localization information tothe sensor node, when needed by the protocolstack or by the application.While all the research on underwater networking so far has followed the traditional layered approach for network design, it is an increasinglyaccepted opinion in the wireless networking community that the improved network efficiency, especially in critical environments, can be obtained witha cross-layer design approach. These techniqueswill entail a joint design of different network functionalities, from modem design to MAC and routing, from channel coding and modulation to sourcecompression and transport layer, with the objectiveto overcome the shortcomings of a layered approach that lacks of information sharing acrossprotocol layers, forcing the network to operate ina suboptimal mode. Hence, while in the followingsections for the sake of clarity we present the challenges associated with underwater sensor networksfollowing the traditional layered approach, we believe that the underwate

summarize the associated physical layer challenges for underwater networking. In Sections 5-9 we dis-cuss physical, data link, network, transport and application layer issues in underwater sensor net-works, respectively. In Section 10 we describe some experimental implementations of underwater sensor networks while in Section 11 we draw the