ultra-wide band sensor networks in oil and gas...

INTRODUCTION

A wireless sensor network is a well-knownparadigm that collects a set of emerging technolo-gies that will have profound effects across a rangeof industrial and scientific applications. Recentdevelopments in wireless technologies and semi-conductor fabrication of battery-powered minia-ture sensors are making the radio devices morecost-effective for a growing number of pervasiveapplications. New technologies such as millimetricwave and Ultra-Wide Band (UWB) [1] are

expected to provide wireless connectivity for highdata-rates within very short range distances.

Although a wide number of applications havebeen proposed for sensor networks, their marketpenetration is still very fragmented with volumesthat are far from the full exploitation of the poten-tials of the latest cutting-edge technologies. Thescientific community should provide innovativetools in all the new cable replacing applications toboost industry toward larger business volumes.

WIRELESS GEOPHONE NETWORKS FORSEISMIC EXPLORATION

The unavoidable need for oil and gas as leadingenergy sources for the next decades is pushingthe oil companies to increase the investments inseismic exploration of new reservoirs and in newtechnologies to improve the quality of depthimaging for more efficient production [2].

Conventional cable-based land explorationoperations (Fig. 1) rely on an acquisition phasethat is based on telemetry cabling to handle remotecontrol commands and to collect data samplesfrom remote sensors (or geophones) in real-time.Seismic crews supervising each acquisition typicallycarry up to one third of a million sensors only forthe operations over one survey. The use of cablingaccounts for up to 50 percent of the total operatingcost of a typical land survey and up to 75 percentof the total equipment weight [4]. Cable-based sys-tems impose such stringent constraints on the sur-vey design, efficiency, and cost that moving towireless architectures is now regarded by oil com-panies as a natural evolution for high-resolutionseismic explorations employing up to one millionsensors within the next decade [3].

Wireless Geophone Network (WGN) is theacronym that will be used in this article to indi-cate the network architecture supporting futurehigh-resolution cable-free seismic explorations.Technical limitations of off-the-shelf wirelesstechnologies force current proposals for WGNarchitectures to choose wireless only for remote

IEEE Communications Magazine • April 2013150 0163-6804/13/$25.00 © 2013 IEEE

ABSTRACT

Seismic exploration and monitoring for oil andgas reservoirs is a peculiar application that requiresa large number (1000–2000 nodes/sqkm) of geo-phone sensors deployed outdoors over large areas(≥ 40 sqkm) to measure backscattered wave fieldsfrom artificial sources. A storage/processing unit orsink node collects the measurements from all thegeophones to obtain an image of the sub-surface.The existing cabling system to connect sensors isknown to cause inefficiencies, large logistic andweight costs, as well as insufficient flexibility in sur-vey design. Oil companies are therefore expectingthat wireless connectivity will provide the enablingtechnology for future seismic explorations. Thisapplication represents a new challenging researcharea for the wireless community. Early results sug-gest that current off-the-shelf radio solutions donot guarantee the minimum requirements in termsof system usability and energy consumption, noteven for current deployment size. This article pre-sents a tutorial view to introduce the basic princi-ples of seismic acquisition systems that arenecessary to define the wireless geophone networkspecifications. Strict sampling synchronization con-straint over large geographic areas, high precisionsensor localization, and high data rate are allrequirements calling for a scalable network systemwhere Ultra-Wide Band radio transmissions play akey role as the only viable technology.

ACCEPTED FROM OPEN CALL

Stefano Savazzi, IEIIT institute of the Italian National Research Council (CNR), Milano

Umberto Spagnolini, Politecnico di Milano, DEIB

Leonardo Goratti, Joint Research Center of the European Commission, Ispra (IT)

Daniele Molteni, Politecnico di Milano, DEIB

Matti Latva-aho, Oulu University, Center for Wireless Communications

Monica Nicoli, Politecnico di Milano, DEIB

Ultra-Wide Band Sensor Networks inOil and Gas Explorations

SAVAZZI LAYOUT_Layout 1 4/1/13 1:31 PM 50

IEEE Communications Magazine • April 2013 151

data quality control while cables are still mostlyadopted for real-time data delivery [3]. Recentadvances in wireless technology have led the sci-entific community to now be mature enough tomeet the rigid constraints imposed by seismicacquisition systems [4].

The main goal of this article is to disclose thespecifics of this fairly unique industrial applica-tion to the wireless community and to proposethe guidelines for a new architecture based onUWB technology. After a brief tutorial on landseismic acquisition systems, the WGN architec-ture is specifically designed according to the net-work requirements. Locating the geophonesthrough self-localization techniques and adopt-ing data compression algorithms to relax therequirements on network throughput are finallydiscussed as two areas that have the potential tolargely improve the survey quality, while openingnew research opportunities.

REVIEW OFLAND SEISMIC ACQUISITION

In land seismic acquisition one (or more) energysource(s) such as dynamite or a controlled-sourcelike vibrated plate or vibroseis are placed on thesurface to generate elastic waves that propagateover the sub-surface. These elastic waves arereflected and refracted by media discontinuities

with different elastic properties (Fig. 1). Back-scattered wavefield is measured by sensors. Ingeophysicist jargon, the seismic channel is usedto refer to the stream of digitalized samplesdrawn from one sensor. The number of seismicchannels depends on the size of the survey.Large-size systems under design are targetingmore than 300K channels simultaneously active,moving to one million within the next decade.

As shown in Fig. 1, the node (or receiver ingeophysicist jargon) might collect one or multiplesensors (if using multi-component receivers). Tosimplify, the sensing operation is similar to a verylarge array of microphones. Sensors can be geo-phones or accelerometers: these are closely cou-pled to the ground to measure the back-scatteredwavefield that conveys reflected elastic energygenerated by seismic sources. After synchronoussampling, the seismic channels are forwarded byeach node to a storage/processing unit to identifythe geological structure of the substrate.

The acquisition system consists of two distinctphases that are repeated periodically (the readermight refer to the wide literature for more in-depth discussion, see e.g., [2]):• The shooting phase where one (or more)

source(s) placed in a predefined position(s)is generating the elastic wave.

• The data delivery phase where the seismicdata is synchronously sampled, quantized,and forwarded by the nodes toward the

Figure 1. Overview of land seismic acquisition system alternating shooting and delivery phases over a sequence of source locations (ontop). Typical 2D WGN network topology and node structure (bottom figure).

Shooting phase

Gas/oil/water

3)

...

Data delivery phase

Gas/oil/water

4)

Shooting phase

Gas/oil/water

ShotTime

1)NodesSource

Data delivery phase

Gas/oil/water

Cro

ss-li

ne d

imen

sion

s

20-4

0m

Shot line

In-line dimension

2)

...Delivery

Time

...

...

Shot

TX/RXNode

Sensor#1

Sensor#M

Source

∆x = ∆y = 5 - 30m

R=48kbps(uncompressed)

Storageunit

Time

...Delivery

Time

SAVAZZI LAYOUT_Layout 1 4/1/13 1:31 PM Page 151

IEEE Communications Magazine • April 2013152

storage unit. Data collection is obtained bymultiplexing data streams from multipleseismic channels.As shown in Fig. 1, shooting and data delivery

are repeated periodically by moving the seismicsource(s) over predefined positions (see the map-view in Fig. 1). The storage/processing unit esti-mates the elastic discontinuities of the sub-surfaceby combining the data received from all the nodes.The Common Shot Gather (CSG) view collectsthe signals recorded by all the available nodes(seismic traces) during one shooting phase and canbe used to visualize the digitalized seismic chan-nels into a two-dimensional space-time grid.

SENSOR TECHNOLOGY:GEOPHONE AND MEMS ACCELEROMETERS

Coil based geophones have been regarded as aproven technology and have been used for a longtime by the seismic industry. Seismic data fromgeophones is obtained by measuring the voltageinduced from the relative velocity of the magnetcompared to the coil. Nodes are typically equippedwith multiple analog geophones to perform electri-cal summation of the signals acquired. For eachnode one single trace is processed, digitalized,stored, and transmitted. Cabling is typically neededto serially connect the geophones to form arrays.

The adoption of micro-machined sensor(MEMS) accelerometers is expected to reduce theequipment weight to improve survey flexibility stillby maintaining the same imaging quality of geo-phone arrays. The expected key advantage ofMEMS accelerometers for seismic exploration is thebroadband linear frequency response compared togeophones that typically extends from DC to 800Hz.MEMS can be printed on a tiny silicon chip withsmaller size (1 cm) and weight (lower than 1 g)compared to coil based geophones (typical weight of70-80 g) [2]. Compared to coil-based geophones, theintroduction of MEMS as a future technology forcable-free surveys will pose more stringent con-straints on battery lifetime of remote units.

To comply with the conventional notation, theterm “geophone” will be used in the following sec-tions to indicate the wireless node device, althoughit is understood that the whole sensor hardwarewill be subject to technological innovations.

HIGH-DENSITY ANDWIDE-AZIMUTH ACQUISITIONS

High node density and wide-azimuth are two keyrequirements in seismic acquisitions. High densi-ty of geophones provides a high-quality depthimaging while wide-azimuth enables the highestresolution of depth images from back-scatteredwavefield of side-view.

Geophones are deployed on the surface toform pseudo-random 2D arrays with an applica-tion-specific deployment as outlined in Fig. 1.The 2D array can be virtually organized intostrips (receiver lines) with nodes ideally placedover rectangular (or rhombic) lattice with hori-zontal (in-line Dx) and vertical (cross-line Dy)spacing (offset) of Dx = Dy = 5–30m.

In practice, natural and man-made obstruc-tions make the network deployment far frombeing regular. Strips can be 10km long (in-line

dimension) and contain thousands of nodes. Thenumber of lines should be large enough to guar-antee wide-azimuth coverage, and this is typical-ly obtained when the cross-line dimension iscomparable with the in-line dimension for moni-toring field extension of 20–100 sqkm. A similarlayout is designed for source deployment toexplore source/geophone reciprocity [2].

SHOOTING-BLIND VS.REAL TIME TELEMETRY SYSTEMS

Local data storage with cable-free nodes sendingonly few bits for data quality control (QC) is clear-ly the most simple alternative (although not fullycompetitive) to conventional wired telemetry. Thecontrol unit is referred to as “blind” since it doesnot receive seismic data while shooting. After syn-chronous sampling, seismic data are stored for thewhole survey duration with time-stamps that labeleach shooting phase and downloaded into theglobal data storage only at the end of the surveywhen geophones are collected back.

Faulty recordings severely impair the wholeacquisition with unacceptable loss of depth reso-lution (and re-shooting after all), so remote QCof seismic data during acquisition is thereforemandatory. QC parameters are transmitted byeach wireless device either periodically or ondemand (Fig. 2). The data-rate for QC informa-tion typically ranges between 1-10 bytes/min andit is one order of magnitude lower compared tothe full seismic data transfer thus relaxing therequirements on network throughput.

Wireless QC shooting-blind systems havestricter requirements and higher security con-cerns compared to conventional wireless sensornetwork applications. First, they should provide alow delay service to guarantee a timely access tomeasurements and control information overlarge-size survey areas. Second, end-to-end datareliability has to be high enough to guarantee anefficient response to any possible alarm condi-tion. Third, the choice of the low-power wirelesstechnology should guarantee the highest reliabili-ty in harsh environments (forests, urban/sub-urban scenarios, etc.) where propagation of radiosignals might suffer from a significant level ofinterference and non-line-of-sight. These require-ments are common to a broader class of closed-loop automatic industrial process control systems.

The IEEE 802.15 Task Group 4e is currentlypushing to enhance the 802.15.4-2011 MAC tobetter support these specific industrial markets.The problem of defining the necessary policiesto support mesh networking to extend networkcoverage is instead the main focus of the IEEE802.15 Task Group 5 [5]. Industrial organiza-tions such as HART, ZigBee, and the Interna-tional Society of Automation (ISA) are alsopushing toward the definition of common speci-fications for wireless industrial process control.It is therefore envisaged that these standards willplay a crucial role for the next generation wire-less seismic equipment in the future.

A real-time wireless telemetry system requiresthe entire seismic data to be transferred duringacquisition (while shooting). This is the solutionthat will be more deeply analyzed in the followingsections as it completely replaces the cabling func-

High node density

and wide-azimuth

are two key require-

ments in seismic

acquisitions.

High density of

geophones provides

a high-quality depth

imaging while

wide-azimuth

enables highest

resolution of depth

images from back-

scattered wavefield

of side-view.



tionalities. Data delivery enables real-time moni-toring of the overall process at the control unit toidentify faulty traces with a great speed up of theoverall acquisition. In addition, by eliminating theneed for local data harvesting, it reduces the risksof data loss caused by equipment malfunctions.

NETWORK REQUIREMENTS FORWIRELESS SEISMIC EXPLORATION

Even if the last decade has witnessed many newproducts and standards in the area of wirelesscommunications, real-time seismic acquisition hasmany peculiarities that are not common to otherwireless sensing networks. A typical survey of300K co-located live-channels is still fairly complexto be fully wireless with the available technology.To simplify, the amount of traffic/connections iscomparable to those that one cellular phone oper-ator handles in a medium size city [4]. In addition,remote units are left unattended for such a longtime that self-organization of the network and lowpower consumption become mandatory require-ments. In what follows we focus on the data deliv-ery phase of the real-time WGN telemetry systemwhere battery-powered nodes are equipped withradio transceivers.

NETWORK THROUGHPUTSome basic data-rate budget analysis is mandato-ry to better capture the throughput requirementsof the network. Given that the minimum seismicwavefield sampling time is Ts = 0.5 ms, the data-rate of one seismic channel generated by onesensor with N = 24 bits/sample A/D is Rc = N/Ts

= 48kb/s with typical value of Rc = 12kb/s forsampling time Ts = 2ms. The aggregated bit-stream generated by one wireless node scaleswith the number of seismic channels that arelocally processed and multiplexed. The type ofsensor and the method by which seismic data isrouted to the storage unit (e.g., by multi-hoptransmission) have major influences on theamount of aggregated data rate.

As an example, aggregated bit-stream fromone node that is multiplexing 2000 seismic chan-nels easily reaches 2000 ¥ Rc = 100Mb/s. A sur-vey deploying 300K seismic channels mightrequire the deployment of 100K tri-componentwireless receivers: this system would need tosupport a sum-throughput of 4.8Gb/s that scalesdown to 1.2Gb/s if using Ts = 2ms. Supportingsuch a high aggregated throughput is quiteunusual for conventional sensor networks, and itrequires the adoption of specific wide-band (ormulti-carrier) radio technologies [1] to improvethe spectral efficiency. In addition, the adoptionof data compression techniques is also mandatoryto relax the data rate requirements.

DEPLOYMENT ANDNETWORK-LAYER MANAGEMENT

Nodes are deployed in pre-defined spots accord-ing to the acquisition geometry outlined earlier.Node placing is done prior to the shooting/deliv-ery phase and in most cases it is a one-timeactivity [4]. Some geophones can be added tothe network while acquisition is in progress, e.g.,to add new lines for enhancing the azimuth cov-erage or replacing faulty sensors.

Figure 2. Shooting blind Quality Control system (QC land system) architecture.

Table 1: Radio technologies for QC land systems

Radio tech.

ZigBee/WirelessHART

WiFi(11abg)

Radio range

100m (omni 5 dBi)

10km (dir. 24dBi)

Transfer rate

250kbps

1Mbps

IEEE 802.15.4closed-loop QC

Sync

(GPS

)

In-line dimension

IEEE 802.15.4/ISA100.11a

mesh network (e.g., 300 nodes) 1-10 byte/min

Storageunit

QC landsystem

Gatewaynode

Long-range WiFi

>10

m

1-4kbit/s

Long range (WiFi)communication

Data delivery enables

real-time monitoring

of the overall process

at the control unit to

identify faulty traces

with a great speed

up of the overall

acquisition.

In addition, by

eliminating the need

for local data har-

vesting, it reduces

the risks of data loss

caused by equip-

ment malfunctions.



The network is mostly static as nodes are notrequired to change their location after their ini-tial deployment. This suggests that the time dur-ing which any two wireless devices can staywithin the communication range is long-enoughto justify protocols supporting hierarchical net-works with self-configuring entities. Channel fad-ing impairing the wireless links might still exhibitlarge-scale fluctuations over time due to environ-mental influences or land surveyors operations.

The expected number of devices required forseismic applications calls for an enormousaddress space. This suggests that the use of IPv6over the low-power wireless network (e.g.,6LoWPAN) might become a candidate option inthe near future. A number of reasons make IPv6also attractive for network-layer management:these include the support of interoperability forheterogeneous networks and the availability of ahuge library of tools and utilities to guaranteeflexibility in many different network conditions.

LOW-POWER OPERATION ANDENERGY HARVESTING

The system needs to be designed to work contin-uously for days (7–30 days). This poses stringentconstraints on the radio transceiver and themedium access control (MAC) policy to pre-serve battery life [6]. In addition, while coilbased geophones require no electrical power tooperate, power consumption for MEMS is in theorder of tens of milli-watts. Today commercialbatteries used for industrial applications have anenergy density below 800 Joule/g with typicalcapacity of 19Ah. However, their performancecan be worse at low-temperatures. Althoughresearch has shown that it is possible to increaseenergy density by tenfold within a few years witha corresponding reduction of battery weight, thebattery capacity typically doubles only every 8-10years. The introduction of energy harvestingtechnologies [12] is therefore expected to be themost relevant breakthrough for next generationwireless land acquisition systems for batteryrecharging or even battery-less devices.

SYNCHRONOUS ACQUISITION ANDLOCALIZATION

Back-scattered elastic wavefield is synchronouslysampled and A/D converted all over the surveyarea. Synchronization in cable-free systemsneeds to distribute the time-reference over thesurvey area with 10–20ms of tolerable samplingskew/jitter. For large surveys (≥ 40 sqkm), timingcan be provided by several master clock refer-ences (e.g., GPS/GALILEO) where each clockdistributor covers an area of 1–3Km radius. Thissolution reduces the jitter accumulation of timereferences as for chain-connected equipment inmultihop/mesh communication.

Accurate node positioning is mandatory toavoid severe degradation of depth imaging.Satellite navigation systems fail to provide local-ization if satellites are not visible such as in jun-gle or forest areas. However, mesh topology canreduce the number of GPS equipped devices byexploiting the concept of cooperative localizationthat can reach an accuracy below 1m.

DATA PROTECTION AND SECURITY

Security algorithms are mandatory to assure thefull protection of the infrastructure functionalityand the wireless seismic equipment. Encryptionmethods should guarantee secure communica-tion between all the wireless equipment to keepthe seismic trace information from being decod-ed by other parties. The advanced encryptionstandard AES-128 complies with the require-ments of seismic acquisition.

WIRELESS GEOPHONE NETWORKS:CLUSTER-MESH ARCHITECTURE

Given the large field extension of the survey, anheterogeneous network that exploits the advan-tages of long and short-range radio technologiesseems to be the natural solution. Figure 3 showsthe hierarchical WGN architecture: the wide sceneis broken into sub-networks managed by a Gate-way node serving as coordinator (and local sink)connected to the storage unit. The Gateway nodeis expected to manage two radio technologies: thelong-range technology is used to connect to thestorage unit while the short-range radio is used tocover the sub-network. The hierarchical architec-ture proposed to comply with the requirements ofoil exploration has the merit of introducing net-work scalability. Sub-networks are organized intoclusters where nodes connect to the associatedcluster-head device. Leaf nodes are the geophonesthat periodically forward seismic data toward theassociated cluster-head, also receiving control com-mands (e.g., acquisition start and stop commands).A cluster-head node might be equipped with sen-sors as with the leaf nodes or be just a relay node.

ULTRA-WIDE BAND TECHNOLOGY FORSEISMIC DATA DELIVERY

The WGN requires tight time synchronization andmandatory self-localization. Given all the con-straints, UWB technology might become the mostreasonable choice at the physical layer. UWB sig-nals have bandwidth larger than 500 MHz (or frac-tional bandwidth larger than 0.2) so that they canprovide a high quality delay/ranging estimation [8]and large data rates to support bursty traffic. UWBsignals are confined in (unlicensed) frequencybands and with stringent emission power spectraldensity limitations so that they can be used forshort-range transmissions in conjunction with other2.4 GHz based wide-band radio technologies with-out paying meaningful cross-interference. UWBtechnology was first proposed for imaging applica-tions in the 1990s, when time domain impulseradio (IR) was the leading PHY layer communica-tion technology. To support advanced high data-rate applications, two UWB technologies haveemerged: the direct-sequence (DS-UWB) and themulti-band (MB-UWB). DS-UWB adopts variablelength spreading codes for binary phase shift key-ing modulation; RAKE receiver is implemented tomitigate multipath fading. In 2007 the EuropeanComputer Manufacturing Association (ECMA)proposed a number of protocol specifications(ECMA-368 “High Rate Ultra Wideband PHYand MAC standard”) that support the transmission

Encryption methods

should guarantee

secure communica-

tion between all the

wireless equipment

to keep the seismic

trace information

from being decoded

by other parties. The

advanced encryption

standard AES-128

complies with the

requirements of seis-

mic acquisition.



of UWB signals through Multi Band OrthogonalFrequency Division Multiplexing (MB-OFDM).High-speed communication supports up to480Mbps that are guaranteed within short-rangesof 5-10m indoor. Extended range to 30m is expect-ed in outdoor line-of-sight environments with datarates scaling down to 53.3Mb/s.

So far, market penetration of UWB technolo-gy for personal area networking is limited whileWiFi standard evolutions have the benefit ofcheaper chips that demonstrated a considerablyhigher appeal to the consumer applications mar-ket. In order to fully exploit the potential of thelatest technologies, we believe the evolutionaryroad for high-throughput UWB should includesensor networks as the secondary (or even pri-mary) target application. In particular, wirelesstelemetry for high-density seismic exploration isone of the candidate applications where the highdata-rate UWB technology can be seen as themost suitable option for PHY.

HIGH DATA-RATE UWB TECHNOLOGY ANDFRAMING STRUCTURE FOR WGN

Given the requirements of the application interms of data-rate and network scalability, time-domain based IR-UWB technology could not be

considered as an option. The complexity of RAKEreceivers and the critical synchronization require-ments currently prevent full-function implementa-tions of DS-UWB technology. MB-OFDM radioscombine the benefits of OFDM transmission withthe network scalability offered by multi-bandtransmission. The MB-OFDM medium accesscontrol is also expected to enable considerablespatial reuse and network scalability through theuse of time-frequency (TFI) and fixed- frequency(FFI) codes to interleave information data overadjacent frequency bands or consecutive OFDMblocks. Although sensitivity to timing errors, phasenoise, and frequency offsets still represent chal-lenging problems for advanced circuit design, theimplementation of MB-OFDM technology is lesscritical compared to DS-UWB. In addition, itreceived wider support from a large number ofmanufacturers and consortiums [7]. MB-OFDM istherefore envisaged as the physical layer technolo-gy for cable-replacing within each sub-network.

Framing structure is taken from the ECMA-368standard with some modifications that are manda-tory for the oil exploration application: these areoutlined in Fig. 4. The Beacon Period (BP) isplaced at the beginning and contains up to 96 Bea-con Slots (BS). BSs are transmitted at every super-frame and carry essential information on device

Figure 3. Real-time telemetry system: wireless geophone network (WGN) cluster-mesh network architecture.

Storageunit

Gatewaynode

WiFi-mesh 5-50Mbps>

10m

2-15Mbit/s

Shot line

Storageunit

6-48 kbit/s(uncompressed)

Leaf node (Geophone)Cluster-head node

Sync

(GPS

)

Gateway node

Cluster-head node

Short range (UWB)comunication

Long range (WiFi)comunication

MB-UWB cluster-mesh network(e.g., 300 nodes)

Real-timetelemetry system

Cluster In-line dimension

The evolutionary

road for high-

throughput UWB

should include sen-

sor networks as the

secondary (or even

primary) target appli-

cation. In particular,

wireless telemetry for

high-density seismic

exploration is one of

the candidate appli-

cations where the

high data-rate UWB

technology can be

seen as the only suit-

able option.



status (neighbour information, data rate, signal-strength), beacon period occupancy information,available and reserved transmission resources. Theremaining part of the super-frame is sub-dividedinto slots (medium access slots — MAS) of 256mseach. Devices can send their information afterreserving collision-free transmission resources interms of one or more MASs, or by using randomaccess within the final part of the super-frame toavoid interference with collision-free transmissions.

The Gateway and the cluster-head nodes areGPS synchronized and issue a unique beacon in areserved BS. The beaconing concept of ECMAguarantees easy-to-spread synchronization, acqui-sition timing and connectivity for large size net-works [7]. In addition, the distributed beaconingconcept constitutes a unique opportunity to sup-port highly dense and large size networks resultingas the interconnection of many clusters of sensors(piconets). The Gateway node behaves as interme-diate sink and periodically issues a unique refer-ence time valid for all the devices within thesub-network and referred to as the beacon periodstart time (BPST). After detecting the BPST, eachcluster-head transmits beacon frames within theBP to maintain and propagate the reference time.

ENERGY-AWARE BEACONING DESIGNLow-energy consumption is a key requirementthat must be specifically addressed for WGN.Since the ECMA specification is not optimized forlow energy [7], amendments to the standard aretherefore recommended by adapting some princi-ples from the IEEE 802.15.4 specifications. ECMAdefines the distributed access to the BP through-out the mechanisms of extension and contractionwhen k devices (k ≥ 2) make random access on n(n £ 8 as ECMA specified value) free slots leftpurposely after the highest occupied beacon slot(HOBS). Since cluster-head nodes have to keep thereceiver active for all the announced length of theBP, unnecessary long and fragmented BPs causeenergy waste from idle listening. Scaling down thenumber of devices that access the BP is mandatoryto reduce the time all the devices should listen forbeacons and minimize the power expenditure forreceiving. To that end, it is proposed to modify theECMA standard to give only the cluster-heads andthe Gateway nodes the right to access the BP, whilethe other (reduced function) leaf nodes can onlyreceive beacons, with no right to occupy any slot inthe BP. Energy consumption per super-frame dur-ing BP listening is quantified in Fig. 4 for differentBP configurations and access schemes. Accordingto ECMA, BP expands when newcomer cluster-head nodes are requesting for beacons: the energyexpenditure per super-frame varies correspondinglyas shown in Fig. 4 (red curves). The proposedscheme for WGN is referred in Fig. 4 as coordinat-ed access (black curve). This scheme allows thecluster-head nodes to compete making first accesson the two signalling slots instead of expanding theBP by randomly choosing a BS after the HOBS.The device that wins the contention is granted theright to transmit a beacon frame using the firstunoccupied beacon slot.

The numerical example illustrated in Fig. 4provides a basic concept of the overall system byassessing the device battery lifetime and the delayexperienced during network set-up. In this case

study it is assumed that each cluster-head is coor-dinating a maximum of 15 leaf nodes. Leaf nodesare randomly deployed over the linear/strip net-work deployment illustrated in Fig. 1 now withspacing that ranges between 5-10m as typical for ahigh-density survey. In each sub-network, thenumber of active cluster- heads aggregating seis-mic data from leaf nodes is 20. Energy consump-tion measurements are taken from a noncommercial MB-OFDM transceiver used as a test-bed [9]. Simulations of distributed beaconing areverified through an OPNET simulator. TheECMA-368 compliant distributed access causes asaturation of the beacon period when a relativelylarge number of devices contend for beacon slots(this is also the case of Fig. 4). On the other hand,the proposed coordinated access is likely to pre-vent this to happen, at least for all the populationvalues that are of interest for oil exploration appli-cation. Compared to distributed access, coordinat-ed access increases the network set-up delay inexchange for larger node lifetime: this is reason-able for static environments where beacon slotassociation remains fixed for a large fraction oftime. Network set-up for beaconing and resourceallocation has a duration of 42 sec for coordinatedaccess while 11 sec are required for ECMA com-pliant access. The energy consumption per super-frame is assumed to remain constant until batterydepletion: this makes it possible to compute theexpected battery life of the cluster-heads that areclosest to the Gateway as these represent the bot-tleneck of the system and typically experience thesmaller lifetime. For a commercial battery withcapacity 19Ah, the worst-case cluster-head lifetimefor coordinated access is 29 days of continuousrecording mode, while it scales down to 14 days forECMA compliant access. Notice that today landacquisitions require continuous recording for 7–10days, while next generation telemetry systems willrequire battery lifetime of several weeks (and up to1–2 months) [15]. To support larger lifetime, addi-tional investigations are required to tune PHYlayer parameters to enable further energy savings.

COOPERATIVE TRANSMISSION WITHGUARANTEED TIME SLOTS

Leaf nodes inside clusters are reduced functiondevices as these communicate to the cluster-head node using random access within the finalpart of the superframe. Each individual geo-phone device collects a relatively small amountof data compared to the capacity of an MB-OFDM radio link (up to Rc = 48kb/s withoutcompression) so that the use of random accessbecomes a reasonable choice. In the process ofsending information to the Gateway node, thecluster-heads must instead process and routelarge volumes of seismic data with typical rangeof 2–15 Mb/s (e.g., to aggregate 300 seismicchannels). Therefore, for cluster-heads the useof a distributed resource reservation policy forrouting data toward the Gateway is mandatory.The cluster-head nodes use the so called dis-tributed reservation protocol (DRP) defined byECMA-368 to route data between themselves.Transmission resources for cluster-heads areassigned in the form of one (or more) guaran-teed time slot(s). Resource allocation can be

The Gateway and

the cluster-head

nodes are GPS

synchronized and

issue a unique

beacon in a reserved

BS. The beaconing

concept of ECMA

guarantees easy-to-

spread synchroniza-

tion, acquisition

timing and connec-

tivity for large size

networks.



granted through the exchange of request/replymessages embedded within the beacon frames.Enabling cooperative transmissions among thecluster-head nodes [10] is also highly recom-mended to guarantee efficient data delivery

toward the Gateway. Cooperative network archi-tectures [11] allow cluster-heads serving as relaysto perform DRP requests for other devices incase transmission resources are not sufficient totransfer all the aggregated data to the Gateway .

Figure 4. Energy consumption (per superframe) assuming 20 cluster-heads joining the WGN. Transmit and receiving current consump-tion values are 200 mA and 350 mA, respectively [9] (voltage of 3.6 V). Cluster-heads are coordinating a maximum of 15 leaf nodesrandomly deployed over the linear/strip network deployment (see top righ-corner sub-figure) with spacing that ranges between 5 and10m. Battery lifetime for cluster-heads closest to the Gateway is computed by assuming: i) 300 active seismic channels per sub-net-work; ii) compression factor N = 5 bit/sample; iii) battery capacity 19Ah; iv) receivers are equipped with geophones. Set up delays are11sec (red curves) and 42sec (black curve) for distributed and coordinated access respectively. Estimated lifetime is 29 days for coordi-nated access while it reduces to 14 days for distributed access. Simulations of distributed beaconing are verified through OPNET simu-lator (dashed curve). Top left-corner sub-figure: framing structure for WGN.

Number of superframes (1 superframe = 65.6 ms)

Energy consumption of coordinated vs. distributed access - 20 cluster-heads

In-line dimension

Super frame

Source

A

B

C

D

Contentionaccess periodslotted CSMA

(CAP)

Guaranteedtime slots (GTS)

DCB

ECMA 368 beacon period (BP)

Gateway nodebeacon slot

Signalling slots

Signalling slots

HOBS (highest occupiedbeacon slot)

0 1 2 tEW

BPST

ABA ... ......

100

2

Ener

gy c

onsu

mpt

ion/

supe

rfrm

ae [

mJ]

0

4

6

8

10

12

14

16

200 300

x20

Network set-updelay (11sec)

Network set-updelay: 42 sec.

Resourceallocation(7 MAS tx/rx)

Resourceallocation

(7 MAS tx/rx)

Expected lifetime(bat. 19Ah) 14 days

Expected lifetime(bat. 19Ah) 29 days

x10

0 1 2

BPST

10 t

400 500 600 700

Coordinated accessDistributed access (ECMA 368)Opnet simulation (ECMA 368)

A

BC

D

B

Gat

eway

A

C

D

Signalling slots

BPSTt0 1 2

x10

Signalling slots

BPSTt0 1 2



GEOPHONE COOPERATIVELOCALIZATION

Geophones stay active in the field for several daysbut can be moved during the acquisition and theyneed to be accurately localized for processing pur-poses, even if a satellite could lack from visibility.Cooperative localization [13] provides a method toexploit the high accuracy ranging of UWB signalsin conjunction with measurements from satellite-based positioning systems. The geophone localiza-tion protocol based on the cooperative frameworkreflects the architecture illustrated in Sect. IV-B

where the leaf nodes (without GPS) are controlledby the anchor nodes (cluster-head nodes withGPS) and are prevented from communicating withother nodes outside the cluster. Cooperative local-ization is adopted for MB-OFDM PHY layer sup-porting a single OFDM band of 528 MHz and byimplementing data communication over standard-compliant FFI channels.

Localization is based on times of arrival(ToA) estimation. Each node computes the dis-tance toward at least three neighboring nodesfrom the estimation of the propagation delay.According to the Cramer Rao Bound (CRB) forToA estimation [13] the range accuracy for delayestimation scales as c/2p÷2SNR ¥ b for the sig-nal-to-noise ratio SNR, effective bandwidth b =500 MHz, and c = 3 ¥ 108 m/s.

The use of cooperative localization techniquescan balance the degradation in localization accura-cy experienced by MB-UWB radios as comparedto IR-UWB. The example in Fig. 5 illustrates thisconcept: it shows the mean squared error (MSE)on the location accuracy [14] for different strate-gies and in terms of the error region around eachnode with 91 percent of confidence. Errors havebeen computed using the CRB for the ideal caseof line-of-sight connections and path-loss exponent3. Clock bias is not considered as round-trip mea-surements are assumed. Three different localiza-tion methods are compared: in the first one (case1) nodes can estimate their positions based on thepropagation delays measured from the GPS-equipped anchors within the visibility range (30min this case). The observed average location accu-racy is 3.7m (case 1, red ellipses) and it is far frombeing acceptable for seismic exploration. Accord-ing to the WGN architecture introduced in Sect.IV-B, the nodes can exploit the opportunity tomake additional ranging measurements with allother leaf nodes belonging to the same cluster: thiscooperative localization approach (case 2, greenellipses) yields an accuracy of approximately 1mand it is suitable for seismic exploration applica-tion. The last case (case 3, blue ellipses) is shownas performance benchmark. In this case leaf nodescan ideally exchange ranging measurements withall the devices, irrespective of which cluster theybelong to [14]. Average accuracy for this fullycooperative localization scheme is now below 0.5m.

COMPRESS AND FORWARD OFSEISMIC DATA

Even if high precision 24 bit A/D acquisition isroutinely adopted to comply with large a dynam-ic of signals, source-coding tailored for the appli-cation largely reduces the overall data-rate to fitinto the throughput limitations of low-powerwireless solutions. The source coding schemeproposed here reduces the data-rate by exploit-ing the correlation of signals since larger correla-tion results in a lower number of bits per sample.To comply with the WGN architecture outlinedearlier, every cluster-head node serving as arelay accumulates the data received from previ-ous cluster-head nodes and compresses theobservation by exploiting the trace-by-trace cor-relation. This compress and forward scheme isillustrated in Fig. 6. According to the multi-hop

Figure 5. Cooperative localization for WGN. MSE region for location accuracywith 91% of confidence for: i) conventional tri-lateration from anchors; ii)cooperative localization within the clusters and iii) fully cooperative localiza-tion. The SNR of the radio link scales with the inter-node distance d as d–a,with a = 3. Average node spacing is 10m.

x [m]

Anchor (cluster-head) node(with known position)

100

10

0

y [m

]

20

30

40

50

20 30 40 50

RMSE = 3.71 m - case 1 (tri-lateration)RMSE = 1.095 m - case 2 (cooperative localization)RMSE = 0.46 m - case 3 (full cooperative localization)



transmission, the new trace measured by the sin-gle-component node is jointly coded with previ-ous ones and compressed. Before applying theincremental encoding, discrete time-warping isimplemented by the cluster-head to align previ-ous traces as described in [15]. The purpose is toexplore the practical benefits arising from thejoint exploitation of the temporal and the spatialcoherence of seismic data that is made availableby the multi-hop radio architecture. Compres-sion is obtained at the price of some additionalcomputational burden for each device.

Being a lossy data compression scheme, apractical example illustrates the limits. At thebottom of Fig. 6 it is shown the so called CSGwith N = 24 bit/sample A/D: vertical axes depictarrival travel times measured by one sensor,while horizontal axis gathers the traces recordedsequentially from all the linearly deployed nodesin a one-shot operation. In this example, eachrelay node collects the data from the local sen-sor(s) and performs an incremental coding basedon the data decoded from the previous devices.Data is forwarded to the Gateway placed halfway

Figure 6. Compress and forward for multi-hop with incremental seismic data encoding (data courtesy of eni exploration & production).Residual error (¥ 100) compared to original data (left image) for N = 3, 5, 8 coded bit/sample (with detailed view of original andcompressed N = 3 bit/sample). Corresponding bit rate per seismic channel for varying coded bit/sample N is shown in the top-rightcorner table (for Ts = 4ms).

3 Km

160 nodes

N[bit/sample]

24

8

5

3

Rc [Kbit/s]Ts = 4ms

6

2

1.3

0.8

Compress and forward

(L+ N)-bit(L)-bit

14

Tim

e [s

]

12

10

8

6

4

2

0

RX TX

Previoustraces

Jointcoding Sensor

New trace24 bit

160 nodes

GatewayCompress and forward (CF) for mesh topology.

Residual error: |CSG (24bit) - CSG (N-bit)| (x100)

N=8bitN=5bitN=3bit



over a line of 320 nodes (single-componentreceivers). As sketched in Fig. 6, every nodereceives L bits/sample accumulated from previoushops and adds N bit/sample for each sample torepresent the additional information drawn fromthe local new trace. The accumulated L+Nbits/sample are forwarded to the next node accord-ing to the multi-hop transmission scheme. Figure6 shows the residual error when the source coderuses N = 3,5,8 bits/sample (instead of uncoded 24bits/sample), with a meaningful reduction of therequired bit rate for every seismic channel Rc (seethe table on the top-right corner of Fig. 6). Areduction of Rc up to 3/24 compared to the casewithout compression can increase the number ofchannels each device is able to multiplex.

Compress and forward for high data rate seis-mic surveys is a new application-specific area ofresearch driven by the need of real-time fullywireless instrumentation.

CONCLUDING REMARKSIn this article we introduced the basic principlesof seismic acquisition systems from the wirelesscommunication perspective. A number ofrequirements/specifications for the physical andMAC layer are provided in order to developdense wireless geophone networks for oil explo-ration. The proposed WGN architecture is basedon a mixture of network technologies that areworking in cooperation to guarantee a large-scale, real-time, synchronous and spatially-densemonitoring system that reliably delivers thesensed data across the wireless network. In theproposed system wireless UWB nodes are simul-taneously sensing, self-localizing and organizinginto a cluster-mesh architecture for data delivery(compress and forward relaying). Gateways for-ward the aggregated traffic from UWB nodes toa central storage unit over long range radio links.

The recent technological advances clearlysuggest that the wireless community is nowbecoming mature enough to develop a fullywireless system ready for the very dense landsurveys expected by the oil exploration and mon-itoring industry within the next few years.

REFERENCES[1] C. Park and T. Rappaport “Short-Range Wireless Com-

munications for Next-Generation Networks: UWB, 60GHz, Millimeter Wave WPAN and ZigBee,” IEEE WirelessCommun., vol. 4, no. 14, Aug. 2007, pp. 70–78.

[2] O. Yilmaz,”Seismic Data Analysis. Processing, Inversionand Interpretation of Seismic Data” vol I and II, StephenM. Doherty Editor, Society of Exploration Geophysicists(SEG), 2001.

[3] B. Peebler, “Looking Ahead to 2020 in the World ofGeophysics,” First Break, vol. 29, Jan. 2011, pp. 31–32.

[4] S. Savazzi and U. Spagnolini, “Wireless Geophone Net-works for High Density Land Acquisitions: Technologiesand Future Potential,” The Leading Edge, Special Sec-tion: Seismic Acquisition, July 2008, pp. 259–262.

[5] M. Lee et al., “Meshing Wireless Personal Area Net-works: Introducing IEEE 802.15.5,” IEEE Commun.Mag., Jan. 2010 pp. 54–61.

[6] V. Raghunathan, S. Ganeriwal, and M. Srivastava,“Emerging Techniques for long lived Wireless SensorNetworks,” IEEE Commun. Mag., Apr. 2006, pp.108–114.

[7] J. del Prado Pavon et al., “The MBOA-WiMedia Specifi-cations for Ultra-Wideband Distributed Networks,” IEEECommun. Mag., June 2006, pp. 128–134.

[8] A. Batra and J. Balakrishnan, “Design of a MultibandOFDM System for Realistic UWB Channel Environ-

ments,” IEEE Trans. Micro. Theory and Tech., vol. 52,no. 9, Sept. 2004, pp. 2123–2137.

[9] L. Goratti et al., “Analysis of Multi-Band UWB Distribut-ed Beaconing Over Fading Channels,” Proc. IEEE Int’l.Conf. Ultra Wideband, Bologna, Italy, Sept. 2011.

[10] S. Savazzi, U. Spagnolini, “Cooperative Fading Regionsfor Decode and Forward Relaying,” IEEE Trans. Info.Theory, vol. 54, no. 11, Nov. 2008.

[11] Z. Sheng et al., “Cooperative Wireless Networks: fromRadio to Network Protocol Designs,” IEEE Commun.Mag., vol. 49, no. 5, May 2011.

[12] J. A. Paradiso and T. Starner, “Energy Scavenging forMobile and Wireless Electronics,” IEEE Pervasive Com-puting, vol. 4, no. 1, Mar. 2005.

[13] S. Gezici et al., “Localization Via Ultra-WidebandRadios: A Look at Positioning Aspects for Future SensorNetworks,” IEEE Signal Proc. Mag., vol. 22, July 2005,pp. 70–84.

[14] M. Nicoli and D. Fontanella, “Fundamental Perfor-mance Limits of TOA-based Cooperative Localization,”Proc. IEEE Int’l. Conf. Commun. Workshops (ICC Work-shops), pp. 1–5, June 2009.

[15] S. Savazzi and U. Spagnolini “Compression and Cod-ing for Cable-Free Acquisition Systems,” Geophysics,vol. 76, no. 5, p. 11, Sept.–Oct. 2011.

BIOGRAPHIESSTEFANO SAVAZZI ([email protected]) received theMSc (2004) and Ph.D. (2008) degree in Information Tech-nology (both with honors) from the Politecnico di Milano.From 2012 he joined the National Research Council (Con-siglio Nazionale delle Ricerche – CNR) as Researcher at theI.E.I.I.T. institute. His current research interests includeadvanced wireless sensor network technologies for indus-trial automation and process control. In 2008 Dr. Savazziwon the Dimitris N. Chorafas Foundation Award for bestPh.D. dissertation.

UMBERTO SPAGNOLINI ([email protected]) gradu-ated in Electronic Engineering (cum laude) from thePolitecnico di Milano in 1987. He has been a Faculty mem-ber of Politecnico di Milano since 1990 where he is nowFull Professor of Telecommunications. He is author of morethan 250 papers and patents in several application areasranging from remote sensing, geophysical methods, com-munication and synchronization for wireless networks.

LEONARDO GORATTI ([email protected])received his M.Sc. in Telecommunications engineering in2002 from the University of Firenze. From 2003 until 2010,he worked at the Center for Wireless CommunicationsOulu-Finland, where he obtained his Ph.D 2011. He is cur-rently working on cognitive radio and spectrum sharingtechniques at the European Joint Research Center of Ispra,Italy. His research interests cover MAC protocols for wire-less personal/body area sensor networks, UWB technologyand mmWave communications. He recently joined theresearch center CREATE-NET in Trento.

DANIELE MOLTENI received the M.Sc. degree (with honors)and the Ph.D. degree in Information Technology fromPolitecnico di Milano, in 2007 and 2011, respectively. Dur-ing 2010 he joined Nokia Siemens Networks (Helsinki, Fin-land) as a visiting researcher. His research interests focuson signal processing topics for wireless communications.He received the Meucci-Marconi (Junior) Award in 2008.

MATTI LATVA-AHO ([email protected]) is a professorat University of Oulu and also Head of Department ofCommunications Engineering. He served as the Director forCentre for Wireless Communications – Oulu during1998–2006. He has been active in the research of CDMAcommunication systems since early 90s’. Nowadays hisresearch interests include future broadband mobile com-munications systems.

MONICA NICOLI ([email protected]) received the M.Sc.degree (cum laude) and the Ph.D. degrees in Telecommuni-cation Engineering from Politecnico di Milano in 1998 and2002, respectively. In 2001 she was visiting researcher atUppsala University, Sweden. She has been Assistant Profes-sor at Politecnico di Milano since 2002. She serves as Asso-ciate Editor for the EURASIP Journal on WirelessCommunications and Networking. Her research interestsare in the area of signal processing and wireless communi-cation systems.

The recent techno-

logical advances

clearly suggest that

the wireless commu-

nity is now becom-

ing mature enough

to develop a fully

wireless system ready

for the very dense

land surveys expect-

ed by the oil

exploration and

monitoring industry

within the next

few years.


ultra-wide band sensor networks in oil and gas...

Documents