dsta horizon 2007

-DH2007_cover_final_a1.FH11 4/4/07 8:33 PM Page 1

-contents.FH11 4/10/07 9:37 PM Page 1

C O N T E N T S

Editorial

Joint Data Link Warfare

Cognitive Radio An Overview and its Potential Benefits

JEWEL M&S Environment for the SAF

Indirect Seawater Cooling and Thermal Storage Systemin Changi Naval Base

Managing Intellectual Property In Procurement Applicable Laws and Policies for MINDEF andthe Local Defence Industry

Safety Culture in the Defence Development and AcquisitionEnvironment

Building the Workplace for a Knowledge Enterprise

The Organisation Compass Enterprise Architecture

Reliability Growth Planning and Analysis of a CombatSystem Using Duane Model and Crow Extended ReliabilityGrowth Model

MASINT: The Intelligence of the Future

2

4

16

28

44

56

66

78

90

104

116

-contents_editorial_r1.FH11 4/10/07 7:53 PM Page 3

E D I T O R I A L

Chan Keng LuckEditor, DSTA HorizonsDirector (Corporate Services)Director (DSTA College) up to 14 January 2007

2

Now into its third issue, DSTA Horizons continues to reflect thewealth of ideas present in our thriving local defence ecosystemand to bring out the spirit of knowledge sharing in the community.The heightened interest in DSTA Horizons, seen from the increasednumber of articles submitted, can only be attributed to the authors,past and present, who have dedicated time and much effort toshare their work. Kudos to all of you!

Attesting to the breadth and depth of knowledge abound, thisissue of DSTA Horizons includes 10 articles. Of these, three havebeen presented at international conferences while one has wonthe Commendation Award in the 2005 CDF Essay Competition.The publication begins with the article Joint Data Link Warfare,which examines the use of data links in enabling joint warfare,and posits a conceptual joint data link architecture to enhancejoint operations.

Cognitive Radio and its Potential Benefits, which explores thefuture in radio systems and wireless communications the CognitiveRadio, follows. Essentially a smart radio that possesses situationalawareness and the ability to adapt to the environment to enhancecommunication, the Cognitive Radio has the potential to bringabout vast benefits in both military and commercial operations.

The third article, JEWEL M&S Environment for the SAF, highlightsone of DSTAs many innovations the Joint Modelling andSimulation Environment for Wargaming and Experimentation Labs(JEWEL). The Joint Battle System, a distributed simulation systembased on JEWEL, is currently used by the SAF Centre for MilitaryExperimentation as a test bed and training platform.

As its name suggests, the fourth article, Indirect Seawater Coolingand Thermal Storage System in Changi Naval Base, depicts anotherof DSTAs solutions to the Singapore Armed Forces (SAF).The engineering innovation is cost-effective, resource and energyefficient, and has received the IES Engineering Achievement Award

-contents_editorial_r1.FH11 4/10/07 7:53 PM Page 4

2000, the BCA Energy Efficient Building Award 2002 and the ASEANEnergy Award 2002.

Less technical but nevertheless relevant and important subjects inthe defence technology community are also featured. DSTA Horizons2007 includes articles such as Managing Intellectual Property inProcurement Applicable Laws and Policies for MINDEF and theLocal Defence Industry, which sheds light on the effective frameworkadopted by the Ministry of Defence (MINDEF) and the SAF tomanage intellectual property rights in defence procurement.

Another article, Safety Culture in the Defence Development andAcquisition Environment, deliberates on how to put in place arobust safety management framework and foster a safety culturewithin the defence community.

DSTA is transforming itself into a knowledge enterprise, in whichour members are empowered through the effective creation,storing, and sharing of knowledge. The seventh article Buildingthe Workplace for a Knowledge Enterprise tracks our journey thusfar, and shares the initiatives that will continue to shape andenhance our work processes.

The eighth article, The Organisation Compass EnterpriseArchitecture, examines the use of Enterprise Architecture to designthe blueprint comprising business processes, data, and the ITinfrastructure to help MINDEF and the SAF realise their businessoperating models.

The ninth article, Reliability Growth Planning and Analysis of aCombat System Using Duane Model and Crow Extended ReliabilityGrowth Model, provides a useful case study for professionalsseeking to apply reliability growth methods for developmenttesting. It describes in detail the application of the Duane Modeland the Crow Extended Reliability Growth Model on acombat system.

With the increasing sophistication of technology, better denialand deception techniques are being used to deny traditional meansof intelligence collection. The last article, MASINT The Intelligenceof the Future, explores the potential of a new discipline Measurement and Signature Intelligence (MASINT) and how itcan be developed to boost intelligence collection capability.

We hope these articles continue to stimulate learning and sharing,and look forward to compiling yet another enriching issue ofDSTA Horizons with contributions from the defencetechnology community.

01_JointData_p06_p11.FH11 4/10/07 9:25 PM Page 1

6


INTRODUCTION

A data link essentially enables two parties tocommunicate messages. Data link is not amodern conception. For example, smoke signalswere used by Native American Indians toconvey messages across a distance. Themechanism for creating the smoke signals isfairly simple: it requires only a fire and blanket.The signal has to be visible and is usuallysituated on top of a hill or mountain.Confidentiality is achieved since the smokesignals are devised privately and areprobably context specific. (This is similar to aone-time pad.)

This simple example reveals some salientcharacteristics of a data link:

a. It must enable collaboration acrossa distance.

b. It must ensure secrecy of information.

c. The message to be conveyed must be cleara n d e a s i l y u n d e r s t o o d b y t h ecommunicating parties.

Implicitly, the data link devised by the NativeAmerican Indians can help maintain the stealthof intended receivers, and therefore help createthe element of surprise in a military context.

Ancient techniques for establishingrudimentary data links were not limited tosmoke signals and fire beacons. Signallingmirrors were also used to convey messages.According to Murray (2004), the emperor wasalerted to Marco Polos arrival in 13th centuryChina through a series of sunlight signalsreflected off mountaintops along his route.

Heliographs, tripod-mounted sunlight-reflecting devices which convey messagescomposed of dashes and dots to a designatedtarget at a distance, were used by the Britishin the North Indian and Afghanistan militarycampaigns in the 19th century.

An account of a modern day data link, Link16, given by Kopp (2004) reveals surprisingsimilarities in the use context of data links.Most of the Link 16 terminals were originallyoperated in a receive-only mode (cf. NativeIndian scouts hiding in a forest) and the signalswere broadcast through an Airborne Warningand Control System (cf. mountaintop fire andsmoke) controlling the fighter aircraft withthe Link 16 terminals.

WHAT HAS CHANGED?

It is not difficult to visualise how data links areused for s i tuat ion awareness andsynchronisation of actions both in ancient andcontemporary times. While data links appearto have evolved over time, they may havechanged fundamentally.

A phenomenon widely associated withinverting the pyramid is occurring. Thismeans that the business of sharing secretmessages is no longer limited to a few. Instead,it can involve a lot more participants. Thisrequires creating capacity in the form ofwireless and wired networked infrastructure.Furthermore, it is insufficient to achieve secrecyonly; integrity and authenticity are nowequally important.

Ancient data links have very little information-carrying capacity. They limit the informationrichness. With advances in sensors, it is nowpossible to digitise the battlespace. Anunambiguous digitised battlespace can beformed with the help of a modern data linkthat enables every object, both friend and foe,to be clearly identified, tracked, and ifnecessary, engaged. (This is not the case in theexamples mentioned earlier.)

The definition and design of a data link hasto capture a wider spectrum of operationalcontexts, functions and processes. It has alsoto be synchronised across different services orunits to achieve a degree of integration andoperational effectiveness. For example, an air

-01_JointData_final.FH11 4/4/07 12:29 PM Page 4

7

strike should not result in fratricide and shouldnot slow the advance of ground mechanisedunits; instead, the air strike should reinforceand enhance the mobility of the mechanisedforces. In order to achieve this, both land andair elements are required to share a commondata link capability. This common capabilityprovides ease of integration of all necessaryforce elements and enables a shorterwarfighting cycle to be achieved. Thus, themodern data link is engineered to support avery rich picture of the battlespace and at thesame time reduce the synchronisation time ofall warfighting elements.

Next, we examine the significance of effectivecross-service (or joint) data links. This isillustrated by using two land-air integratedwarfare scenarios: German Blitzkrieg in WorldWar II and Operation Iraqi Freedom (OIF).

GERMAN BLITZKRIEG

Blitzkrieg is German for lightning war orflash war. The concept revolves around thecoordination of tanks, air power, and artilleryin a concerted effort to breach an adversarysline of defence. It is believed that a rapidbreach followed by penetration deep into theenemys rear, destroying logistics and vitalcommand and control centres, should disruptthe enemys battle rhythm. The ensuing shockand surprise would then provide the conditionsfor encircling the enemy forces and crushingthem. This is illustrated in Figure 1.

The ability to coordinate land and air forcesfor the breach and subsequent rapid advanceproved crucial but difficult. Historians havecredited Heinz Guderian with developing thesolution to this problem by equipping tanks,artillery and the air force with High Frequency(HF) radio equipment for communications. Asdescribed by Fiedler (2004), Guderian hadworked out techniques to leverage the NearVertical Incidence Sky-wave mode of HFpropagation. This enabled the German groundforces to communicate over a large area ofoperations on the halt and on the move, aswell as with the air force. The FuG-10 HF radiowith both monopole and loop antennas wasthe mainstay of German Bl itzkriegcommunications.

Tactics evolved to support the ground to aircoordination. Air liaison detachments weredeployed to the ground forces to pass requestsfrom the ground to the air and receivereconnaissance reports. This was an early formof close-air support (CAS).

Although a rudimentary CAS system wasestablished, the Germans did not train to guideaircraft onto the targets. Furthermore, not alltanks were equipped with HF radios. Only thecommand tanks had both the ability to transmitand receive. The fleet comprised mainly HFradios in receive mode. Thus, the CASnetwork was primarily a Command andControl (C2) net. Situation awareness wasmainly achieved through voice.

Figure 1. (Left) Tanks breaching an enemy line with support from dive bombers (GermanStukas). (Right) Forces penetrate deep into enemy rear to destroy bases of support.

(The Origin of Blitzkrieg - WWI, n.d.)


8

The new concept, tactics and technology forproviding ground-to-air coordination provedsuccessful in the invasion of France in 1940 andup to the early stages of the German invasionof Russia in 1941. Despite the innovation oftactics and technology, many problemscontinued to plague combined air and groundoperations. The ground officers saw air supportas a means to conduct mass fire at criticalpoints, overcoming the lack of artillery; thesmaller aerial bombs used also meant thatroads and other transport infrastructure wouldbe left fairly intact, a condition necessary forcontinued force progression into the enemysrear. The air officers contended thatdistinguishing friend from foe would bedifficult, and furthermore, targets on theground engaged in combat would be dispersedand concealed, diminishing the effects of airfirepower (Close Air Support, n.d.).

Essentially, both air and land elements lackeda common recognised ground picture. This wascompounded by a lack of an effective datalink. For example, the Germans had to markthe ground with symbols to signal to air fighter-bombers how far the ground forces hadprogressed. Evidently, the measure wasintended to prevent fraticides, reinforcing theobservation that there was a lack of situationawareness and effective data link capabilities.

OPERATION IRAQIFREEDOM

Fast forward 60 years later. The battlegroundhas changed from Europe to the Middle East.The US-led coalition force has launched OIF.Unlike the Persian Gulf War and OperationEnduring Freedom in Afghanistan, the US didnot launch a lengthy aerial bombardment andamass significant forces before launching theground campaign. Instead, General TommyFranks envisaged a coordinated, simultaneousland-air campaign. Following the Blitzkriegconcept, the US forces would bypass the majorcities and avoid fighting Iraqi Military Units.The centre of gravity was the capture ofBaghdad. Capturing Baghdad would deal aheavy psychological blow to the morale ofIraqi military resistance.

Another objective was to minimise collateraldamage to facilitate post-war reconstruction,especially to the economic infrastructure ofIraq, such as the oil rigs and wells. The campaignimperative was speed and the means tofacilitate rapid ground manoeuvre was throughbattlespace shaping, i.e., Corp CAS. In particular,as V Corp lacked the artillery pieces to supportdivision battlespace shaping, i.e., MultipleLaunch Rocket System (MLRS) to suppressenemy air defences, Central Command madethe decision to distribute Air sorties to V Corpthrough the Coalition Forces Land ComponentCommander (Kirkpatrick, 2003).

The US identified three types of CAS:

Type 1: The controller can see both the targetand the aircraft and directs the aircraft attackon the target.

Type 2: The controller can neither see the targetnor the aircraft but directs the attack on thetarget through intel l igence inputs.

Type 3: Same as Type 2 but occurs in a situationwhere it is assessed to have a low riskprobability of fratricide.

For Type 1 CAS, a simple point-to-pointcommunications capability may suffice.Type 2 and 3 CAS require shared situationawareness and common data link capabilities.In OIF, only six percent (Kirkpatrick, 2003) ofCAS was Type 1. This demonstrates the criticalrequirement all three elements - theintelligence input, such as from UnmannedAerial Vehicles (UAVs), the Joint Tactical AirController (JTAC) and the engaging aircraft -have the same situation awareness of the targetand its environment.

The challenge of CAS was further compoundedby two other factors:

a. The battlespace was non-linear and CASwas required in killboxes that were closeddue to the presence of friendly forces.

b. After Operation Desert Storm, the Iraqisstudied American tactics and adoptedasymmetric strategies to reduce the qualitativeedge of superior American technology. Theywould disperse into smaller units and seek


9


concealment in vegetation and urban areas,changing location frequently every fourto eight hours, usually in bad weather orin darkness.

These factors meant that the blue force hadonly a short window of opportunity to engagethe enemy and it had to do so in the presenceof its own forces without fraticides. In the past,this could mean withholding action andengaging only when coordination with ownforces was achieved. Now, the forces couldoperate in a self-synchronised manner.

Some evidence shows that a data linkedenvironment was conducive for CAS despitethe challenges. For example, just after 10 days,the Medina Division, reinforced by theHammurabi Republican Guard Division, wasreduced to an assessed strength of 29 percentfrom an initial assessed strength of 96 percent(Kirkpatrick, 2003).

Urban CAS also achieved impressive results. Bythe end of the war, urban CAS missions haddestroyed 105 bunkers, 225 buildings and 226targets which included aircraft, command postsand mobile C2 equipment.

OPERATION IRAQIFREEDOM CASVS GERMAN BLITZKRIEG

Proponents of land-air integration such asGuderian believed that effective data link isthe key to land-air integration. The results ofOIF have vindicated this belief. It is also clearthat superior air power led the Iraqi militaryto adopt asymmetric strategies, making thempotentially more difficult to engage as targets.In this aspect, officers of the Luftwaffe wereright. Even with precision weaponry, which issimilar to the idea of using smaller aerial bombsfor CAS in WWII, it would not overcomeasymmetric strategies targeted to avoid thebrunt of airpower.

The ability to detect targets using UAVs andother real-time and near real-time intelligencesources was crucial to bringing precisionweaponry to bear. In urban CAS, delayed fused

Joint Direct Attack Munitions enabled targetsto be attacked with low collateral damage.Such attacks were carried out after sensorssuch as UAVs had detected and tracked thetargets, and sometimes decisions were madeto defer a strike to reduce fratricides, collateraldamage and civil ian casualties. Thiscircumvented asymmetric strategies such asdispersion and concealment.

Closing sensor to shooter loops with precisionand with rapidity differentiated the OIF CASfrom German Blitzkrieg land-air coordination.OIF CAS was precise because all objects weredigitised and de-conflicted before anengagement. This was a result of real-timeblue and red force tracking through a myriadof sensors integrated using near real-timedata links.

The sensor to shooter loops were shorterbecause the sensors and shooters were tightlyintegrated through data links in many cases.A combination of video, situation awarenessand C2 data links created a real-timecollaboration environment for prosecutingType II and III CAS targets. Usually, theenvironment is highly localised and supportsa few nodes (a sensor-C2-shooter system),reflecting a near optimal pairing of sensorsand weapons to target.

In effect, airpower was reinforced; theintegration of a myriad of sensors, intelanalysts, planners and decision-makers usingdata links reinforced the effective use ofair power.

In retrospect, the Germans inability to directaircraft on targets in WWII was a critical gapin joint land-air integration capabilities. OIFappears to have closed this gap admirably(when the weather was good) with its systemof data links. This is indeed a tribute to60 years of remarkable progress made indata links.


10

every five to 10 minutes. The information fromBFT could not be used to gauge the accurateposition of a unit to avoid fratricides.

The video downlinks from UAVs and videopods gave JTACs and SOFs a positiveidentification of a target. It also enabled theJTACs and SOFs to guide shooters such as F16sand F15s to the targets. Battle DamageAssessment (BDA) through the video downlinkswas also instantaneous, enabling a fasterdecision to re-strike if necessary.

While EPLRS and video data links provided thecapability for precise and rapid decentralisedexecution, the suite of CDLs, TCDLs and Link16 provided the means for integratingintelligence with C2, enabling centralisedcontrol and the efficient allocation of resources,such as weapons and sensors to target pairing.

The increase in the sophistication of data linksenabled integration and operationaleffectiveness. However, it came with a price.The Germans in WWII relied mainly on HFradios for communication. This entailed thesame frequency and waveform operations. Theadvantage was greater manageability andunquestioned interoperability across land-airelements. OIF used different data link types.Consequently, for interoperability, gatewayswere required. For example, integrating Link16 and EPLRS/SADL required the TransparentMulti-Platform Gateway (TMPG). The TMPGtranslated Link 16 TADIL-J messages to SADLtype messages.

The different types of data links have createda situation where gateways have become anecessity. The gateways can be deployed onground platforms, such as the BUG-E, orairborne platforms such as the KC135 ROBE.The manageability, mobility, persistence andsurvivability of these platforms have to befactored into the mission equations for success.Recognising the problem, the US has embarkedon the Joint Tactical Radio System (JTRS). Akey piece of technology is the WidebandNetworking Waveform (WNW). This will be acommon capability to enable cross-serviceintegration. The JTRS will harmonise the

OPERATION IRAQIFREEDOM DATA LINKS

The OIF data links deployed were a system ofmultiple data links. Unlike the German systemin WWII, it was not simplistic voicecommunications over HF. It depended on anetworking of a handful of data links. Someof these are:

a. Common Data Links (CDLs) used for downlinking sensor information to Ground ControlStations and C2 nodes. They are used to supportexchange of Intelligence SurveillanceReconnaissance (ISR) information and employedmainly on manned ISR platforms.

b. Tactical Common Data Links (TCDLs), a partof the CDL family, used to equip unmannedplatforms such as UAVs used in ISR applications.

c. Link 16 which is the US Department ofDefenses primary Tactical Data InformationLink (TADIL) based on J-series messages. It isused for C2 messages and air-to-air assets i.e.,surveillance tracks, Electronic Warfare, weaponcoordination, etc. It supports a wide area ofoperations (300 nm diameter).

d. The Enhanced Position Location ReportingSystem (EPLRS) / Situation Awareness Data Link(SADL) which complements Link 16 by providingthe ground situation awareness picture. Aircraftequipped with SADL can also share airsurveillance tracks and C2.

e. Video Data Links which provide information/ video downlinks from UAVs and Litening Pods(on F16s and F15s) to specific JTACs and/orSpecial Operations Forces (SOFs) withspecial receivers.

Of the four data links, EPLRS played animportant role for CAS because it was able todisplay the five closest friendly units withinproximity, regardless of the target position.This was critical as the Blue Force Tracker (BFT)provided a non real-time update of Blue Forcesposition with refreshes occurring approximately

01_JointData_p06_p11.FH11 4/10/07 9:25 PM Page 2

11

incompatibilities and resolve some of the issues.However, because the JTRS programme hasplaced too much emphasis on legacywaveforms, backward compatibility andreplacement of legacy radio systems, it isexperiencing cost overrun and schedule slips.

The programme has to be structured aroundthe WNW but with interoperability with fewerlegacy waveforms. This also means thatgateways will be necessary in the interim tooperate with legacy radio systems.

A multi-data link environment would seeminescapable. With the current emphasis onjoint interoperability, this means gateways onthe ground, in the air and on ships. Thesegateways are a current reality but they do notpromote agility and nimbleness. Very often,they have a large footprint, i.e., US ROBE andBUG-E which is counter to mobility, low-signature and survivability

Should cross-service integration comprise asystem of multiple data links interconnectedby gateways? Should it be one with greatercommonality, and therefore, reducing thenumber of gateways? Which is more efficientand effective?

AN EFFECTIVEJOINT DATA LINKARCHITECTURE CONCEPT

The land-air integration is but one example ofcross-service, integrated operations. In thenomenclature of network centric architecturegiven by Dekker (2005, Architecture D), thejoint architecture is the most complex fordata link solutioning. This is due to the complexmatrix of information exchange requirementsamong platforms and also due to theheterogeneity of the platforms, which imposesdifferent requirements on the data links, i.e.,terrain, speed etc.

Furthermore, it is expected that the jointnetwork-centric architecture will have a richerinterconnectivity matrix as more unmannedsensors and weapons which are harmonised

with manned systems are introduced . Thisintegrated system of systems, conceived toshorten the engagement cycle, will create newdemands on data links in terms of bandwidth,latency and range.

These systems will initially be few, as they haveto undergo a phased transit ion ofexperimentation, integration and operationaltransformation before they are cost-efficientfor mass adoption. This process leads to highdemands for such assets and their products,i.e., sensor imagery, video, etc. Joint access tosuch products, whether in their raw orprocessed form, will be vital to creating anaccurate common operational / tactical picture.The access has to be sustained even while onthe move without limiting the operationstempo. This leads inevitably to the developmentand research of ad hoc networks. In thisapproach, the proponents frame the problemof data links around ad hoc, mobilenetworking capabilities.

An example of this approach is given bythe US Wideband Networking Waveformwhich is currently developed to provideadvanced ad hoc networking capabilities fora joint networking, data-linked environment.Most proponents of ad hoc networkingenvision a network of seamlessly connectednodes with multi-hop capabilities andautonomous routing scaling to thousands ofnodes. In examining the conditions of jointconnectivities, an ad hoc networking capabilityis necessary but not sufficient to achieveeffective and optimal data link capabilities.We may draw some insights from scale-freenetwork research.

In 1998, physicist Albert-Laszlo Barabasimapped the World Wide Web using a webcrawler (Scale-Free Network, n.d.). He foundthat the Web did not resemble a random,distributed network, against conventionalwisdom. Instead, the web exhibited many well-connected nodes. Unlike random, distributednetworks, the proportion of well-connectednodes does not diminish as more nodes areadded, but rather, remains constant. Barabasicoined the term scale-free networks to



12

represent this specific class of networks.It turned out that scale-free networks describea fairly large spectrum of networks,including power grids, social networks andgene-to-gene interactions.

Several properties of scale-free networksinterest us:

a. Scale-free networks are more resilient toerrors than random, distributed networks.Since data transfers in a wireless medium aresensitive to errors (more errors lead to lowerdata throughput), scale-free networks havean advantage.

b. Although scale-free networks are resilientto random errors, they are vulnerable to directattacks on the hubs.

c. Research indicates that removing a high-capacity, direct link to a hub reduces thevalue of the network more than removing longrange but lower-capacity links.

The author has performed some preliminarystudies on adapting the ideas of scale-freenetworks to wireless ad hoc networks (Chia,Tri & Su, 2006). One possibility is to adoptdiversity techniques to reduce the vulnerabilityof direct hub attacks. Thus, in theory, a scale-free network could be made robust to bothrandom errors and deliberate attacks.

It must be emphasised that scale-free networksare different from ad hoc networks in the sensethat ad hoc networks assume randomdistribution of the nodes. To communicate

end-to-end, ad hoc networks do not limit thenumber of hops to achieve communication.In fact, ad hoc networks are designed to beefficient for end-to-end connections acrossmultiple hops. Scale-free network research hasshown that this is not desirable. Instead, directconnections with very limited multiple hopconnections are preferred as this is shown tobe topologically more stable (from the viewpoint of errors).

Dekker (2005) has performed simulations tocompare the performance differences betweenad hoc, random networks with those of scale-free networks, with emphasis on militarycontext. The results are shown in Figure 2.

The parameter p is the Kawachi processparameter that determines the attachmentbehaviour of nodes in large networks. For pequal to unity, a random or distributed networkis formed. (This is usually the ad hoc networkingcase.) For p greater than or equal to two, ascale-free network is produced. (For these cases,the networks have mostly direct connectionsnot with each other, but communicate throughhubs that have evolved to be optimal forconnections.) It is shown in Figure 2 that theperformance, determined by loss exchangeratios of two opposing forces, is better for thescale-free network compared to a random,fully distributed network (fully ad hoc).

The reason for the difference in performanceis two-fold. First, the presence of hubs inscale-free networks facilitates direct links.Second, in the scenario, the links simulated arefairly high-speed links (relative to mobility).

Figure 2. Performance Score of Scale-Free Networks (Dekker, 2005)

0 0.02 0.05 0.1 0.2 0.5 1 2

Average Distance (D) 4.14 3.67 3.42 3.17 2.84 2.61 2.55 2.44

Clustering Coefficient (C) 0.50 0.47 0.44 0.40 0.32 0.20 0.17 0.23

Node Connectivity (K) 4.00 2.99 2.78 2.50 1.93 1.20 1.01 1.00

Symmetry Ratio (r) 1.56 3.32 3.70 4.06 4.56 4.74 4.65 4.48

Performance Score (S) 0.842 0.850 0.863 0.882 0.903 0.897 0.904 0.924

Number of Hubs 0 0 0 0 0.01 0.14 1.33 3.39

Value of Kawachi Process Parameter p


13


The results indicate that ad hoc networking isa necessary condition but not sufficient foroptimal connections of joint data links. Specialalgorithms that mimic the preferentialattachment scheme of scale-free networksshould be incorporated into ad hoc networks.If this condition is fulfilled, it is possible toachieve fault-tolerant, optimal ad hoc networksfor joint data link operations. The results alsomean that it may be more productive to focuson the creation of highly dynamic subnets withpredominantly direct connections and possiblyvery few hubs of opportunity. In this case, adhoc routing is not a dominant consideration.The ability to carefully select these hubsthrough algorithms is the key. Since these hubscarry the bulk of information, the subnetsshould preferably be broadband enabled.

The hubs themselves should then be ad hocnetworked across a very thin backbone,suggesting that this backbone may notnecessarily be itself broadband. The reason forthis is that long range links have to be morerobust and a trade-off between robustnessand data carrying capacity is necessary. Thisconcept is depicted in Figure 3.

Here is an explanation of the differentcomponents:

a. The thin backbone provides wide areacoverage and trades off high data rates for

longer range and wider coverage. It is alsorobust to possible interference.

b. The broadband local area network (LAN) isdevised to be an ad hoc, peer-to-peer networkso that the ability to form hubs temporallyand spatially is facilitated. Furthermore, thelinks should be high capacity in nature; thehigher capacity links ensure that local forceelements are synchronised and function muchfaster to prevent attacks that could inflict harmon the system.

c. Both the thin backbone and broadbandLAN can function independently of each other.However, should the need arise to leverageeach other for extended and expandedsituation awareness, common protocols andmessage translators can provide the means toconnect seamlessly.

The thin backbone is accessible only to aselect group of participants the hubs. Thisensures low latency across very wide coverageand long ranges. Since this is a select group,there are fairly few participants. Second, thethin backbone has to be fairly flexible fortechnology insertion. This provides thepossibility to remove the trade-off betweendata rate and range. In addition, newparticipants can be added. A thin backbonecomposed of software defined radios providing

Figure 3. Joint Data Link Architecture Concept

BROADBANDLAN

BROADBANDLAN

BROADBANDLAN

THIN BACKBONE (WIDE AREA COVERAGE)

NODES CONNECT DIRECTLY(FULL-MESHED)

Translators Protocols


14

flexibility of waveform and frequency shouldbe optimal.

The broadband LAN shall increasingly compriseCommercial Off-The-Shelf equipment such asWiFi and WiMax, and be based on InternetProtocol adapted for military environment,frequencies and range. This creates thecondition for maximum number of participantsto be equipped with the same communicationsmeans. With mass proliferation, peer-to-peerconnections and the exchange of informationare assured. This is the commercial model ubiquity drives connectivity.

In our architecture, hubs are gatewayequivalents. Thus, our architecture alsoadvocates more commonality and less gatewaysfor optimal operations.

CONCLUSION

There is currently no analytical template tocompare different data link architectures forefficiency and effectiveness. Reasons foradopting any data link architecture are alsonot purely driven by technical merits alone.They have a lot to do with the alliances andhence, interoperability requirements forcoalition operations. European countries, forexample, have adopted two primary data links Link 16 and Link 11 (22). This is to inter-operate with US forces and equipment.A second driver is legacy systems. Manycountries have indigenous data links, i.e.,Swedens Ra-90, which are already in-service.These systems have to be taken intoconsideration when developing a data linkarchitecture for joint operations.

Thus, we expect the above conceptualarchitecture, driven purely by technical factors,to evolve taking legacy systems and futuredefined coalition operations into account. Aresultant architecture may then be differentfrom what we have envisioned.

Another consideration is that any adversarywill adapt to superior technologies usingasymmetric strategies. From WWII to OIF andthe Israeli-Hezbollah war, it has been proven

that when faced with an overwhelming force,the adversary will disperse and conceal itself.Cordesman (2006) noted the limitations ofintelligence, target and BDA against anadaptive enemy. Taylor (2005) made the sameobservation that current light ground andaerial surveillance is insufficient to gatherintelligence on an adversary that adapts,disperses and conceals himself using knowledgeof the surveillance capabilities of the US andother Western Countries. The US Army appearsto have evolved its warfighting system to fighta dispersed enemy in line with the observationsof Cordesman and Taylor long before theAfghan and Iraqi wars. The warfighting systemis known as Future Combat System.

Because the adversary is intelligent andadaptive, and we should not assume otherwise,the data link system for joint operations mustbe designed to be capable of supporting futureoperations against fleet-footed, highly stealthyadversaries. Against such adversaries, the datalink must support equally fleet-footed andstealthy sensors, i.e., soldiers on the ground.To be clear, it does not help to give a soldiera Link 16 terminal because of the form and fit.Thus, future data links must be sizedappropriately for disadvantaged users or nodes,i.e., tactical unmanned systems. In this aspect,our architecture which advocates very fewhubs (resource-rich) and many nodes (resource-poor) directly connected with these hubs strikesa good balance between performance,affordability and flexibility.


BIOGRAPHY

15

REFERENCES

Chia, K. B., Tri, T. H, & Su, W. (2006). SimulationOf Physical And Media Access Control (MAC)Layer For Scalable Wireless Sensor Networks.NPS Thesis.

C l o s e A i r S u p p o r t . ( n . d . ) .Retr ieved 15 October 2006 f romhttp://en.wikipedia.org/wiki/Close_air_support

Cordesman, A. H. (2006). Preliminary Lessonsof the Israeli-Hezbollah War. Centre forStrategic and International Studies.

Dekker, A. (2005, November). A Taxonomy ofNetwork Centric Warfare Architectures.Presented at the SETE 2005 SystemsEngineering/Test and Evaluation Conference,Brisbane, Australia.

Dekker, A. (2005, December). NetworkTopology and Military Performance. Presentedat the MODSIM 2005 International Congresson Modelling and Simulation, Modelling andSimulation Society of Australia and NewZealand.

Fiedler, D. M. (2004, March). HF Combat NetRadio Lesson Learned Again. In ArmyCommunicator.

Dr Yeoh Lean Weng is Director (C4I Development) and concurrently, DeputyDirector of Temasek Defence Systems Institute at the National University ofSingapore (NUS). He received his Bachelor (with Honours) and Master of Sciencedegrees from NUS in 1983 and 1987 respectively. He further obtained twoMasters (Distinction) in 1990 and a PhD in Electrical Engineering in 1997 fromthe Naval Postgraduate School (NPS). He attended the Program for ManagementDevelopment from Harvard University in 2003. He received the National DayPublic Administration Medal (Bronze) in 2001, and the Defence TechnologyPrize in 1992 and 2004. Lean Weng has extensive experience in the developmentand implementation of command, control and communications systems.

Daniel Chia Kim Boon is Programme Manager (DSTA Masterplanning andSystems Architecting). He has been involved in the acquisition and R&D planningof communications and data link systems. He is currently responsible for datalink system of systems architecting. He has a Master of Engineering and aBachelor of Engineering in Electrical Engineering from NUS. He obtained aMaster of Science (Electrical Engineering), specialising in communications, fromthe NPS in 2006.

Grant, G. (2006, June). US Army May Scale BackUAV Plans. In C4ISR Journal.

Kirkpatrick, C. E. (2003). Joint Fires as TheyWere Meant To Be: V Corps and the 4th AirSupport Operations Group During OperationsIraqi Freedom. Presented at AUSA 2004.

Kopp, C. (2004, March). NCW Buzzwords, Bytes,and The Battlespace. In Defence Today.

Murray, M. G. (2004). Aimable Air/Sea RescueSignal Mirrors. The Bent of Tau Beta Pi.Retrieved on 4 December 2006 fromw w w. t b p . o r g / p a g e s / P u b l i c a t i o n s /BENTFeatures/F04Murray.pdf

Scale Free Network. (n.d.) Retrieved 15th Oct06from en.wikipedia.org/wiki /Scale-free_network.

Taylor C. D. (2005). Trading the Sabre forStealth: Can Surveillance Technology ReplaceTraditional Aggressive Reconnaissance.Presented at AUSA 2005.

The Origins Of Blitzkrieg WWI. (n.d.).Re t r i eved 15 Oc tober 2006 f romwww.bellum.nu/basics/concepts/blitzkrieg.html


-02_CognitiveRadio_final.FH11 4/4/07 12:59 PM Page 3

18


WHAT IS A COGNITIVE RADIO?

Rene Descartes famously proclaimed Cogito,

ergo sum1 or I think, therefore I am, thereby

convincing himself of his own existence.

Likewise, cognition is the cornerstone of a

Cognitive Radio (CR).

CR is usually discussed in the same breath as

Software Defined Radio (SDR)2, simply because

the best brain is helpless unless there is a body

for it to act with. The SDR is the body which

the CR acts with to sense the world around it

and influence its position in the world. The CR

can thus be viewed as the next quantum step

beyond Software Radio.

CR is a smart radio that possesses self-

awareness, environmental / contextual

awareness, and adapts itself accordingly. The

term Cognitive Radio was coined by Joseph

Mitola III in his PhD dissertation at the

Royal Institute of Technology, Sweden,

and encapsulated the radio ideal

a flexible, multipurpose device which

offers communicat ion everywhere ,

anywhere, anytime.

Like any thinking organism3 the CR has to gothrough a thought process to lead to changesin its internal model (learning) and its actions(adaptation). The CRs cognition cycle is shownin Figure 1.

INTRODUCTION

[Scenario 1 Present Day]

An emergency has occurred at the Suntec

Singapore International Convention and

Exhibition Centre during a mega event where

there are more than 50,000 delegates and

visitors present. There is chaos and panic.

Anxious citizens frantically attempt to use

their mobile phones to contact their loved

ones but the lines are jammed. The entire

homeland security force, including the police

and the military, quickly converges on the

scene. Command Posts are set up

immediately in the vicinity. Attempts to

establish clear lines of communication for the

various units from the homeland security

force have been frustrating, as the

communication devices are either not

interoperable or are jamming one another.

It is a nightmarish situation, since without

any means to communicate, the entire

command and control hierarchy is rendered

ineffective. The scene is one of total anarchy.

The situation stabilises only after a long delay

during which the authorities establish

communication lines via careful spectrum

management and network access regulation.

[Scenario 2 The Future]

An emergency has occurred at the Suntec

Singapore International Convention and

Exhibition Centre during a mega event where

there are more than 50,000 delegates and

visitors present. There is initial chaos and

panic. The entire homeland security force,

including the police and the military, quickly

converges on the scene. Command Posts are

set up immediately in the vicinity. Inside the

centre, anxious citizens use their cognitive-

enabled mobile phones to contact their loved

ones, and despite the heavy call traffic, there

is good call quality and coverage. Outside

the centre, various security units use their

cognitive software-defined handheld radios

to establish clear lines of communication to

their command post and to one another. The

situation stabilises quickly, as the homeland

security forces entire command and control

hierarchy is firmly put in place. The citizens

inside the centre wait confidently for the

situation to be taken care of, with access to

real-time public news via their 4G video mobile

phones.


19

Cognition cycle behaviour can be generalisedinto two main tracks:

Reaction Loop (External loop in Figure 1) The envisioned CR observes the outside world,orients itself in the context of this world, andeither plans, decides, or acts immediately,depending on the urgency of the situation.

Learning Loop (Internal loop in Figure 1) The CR observes the outside world and appliesthe external stimulus to its knowledge basewith artificial intelligence.

Joseph Mitola classified the tasks to beperformed into nine levels based on the levelof cognition required (Table 1).

Until recently, the radio has been a dumbpiece of equipment that could only functionas it had been pre-programmed to, and adaptin a very limited manner when the operatormanually controlled the mode change. Thiswas in a very large part due to the radiosdependence on specific hardware, resulting inincompatible systems which could onlyinteroperate with each other via implementingspecific gateways for each pair of radio typesthat needed to communicate with each other.The gateway was thus the single point of failurein connecting two different radio nets, and

was a critical weakness in force intra-connectivity. In addition, implementing adifferent gateway for each pair of radio typeswould be expensive and laborious for forceswith many different radio types. This is becausethe number of gateways increases dramaticallyas the number of radio types increases.

In the late 1990s, the US embarked on theJoint Tactical Radio Systems (JTRS) programmeto solve its blue force intra-operability problem.The JTRS would replace legacy radios with afamily of interoperable, digital, modular,software-defined radios that operate as nodesin a network to ensure secure wirelesscommunications and networking services formobile and fixed forces (Joint ProgramExecutive Office, 2006). This ambitious goalwas to be realised by the SDR, which bydecoupling the radio from its hardware, wasa radio that could be reconfigured via softwareto perform different functions. This wouldresult in a flexible system and break the thenstovepiped development of radio systems.

Thus, the SDR marked the first evolutionarystep towards the fully adaptive radio idealisedby the CR.

Figure 1. Cognition Cycle (Mitola, 1999)

Generate Alternatives(Programme Generation)Evaluate Alternatives

-02_CognitiveRadio_p20_p22.FH11 4/10/07 9:28 PM Page 1

20

WHY IS COGNITIVE RADIO VIEWED AS A DISRUPTIVE TECHNOLOGY 4

The SDR can be described as todays front-runner in radio research and technology, andyet, cannot truly claim to have attainedcognition level 1. The SDR, for all its flexibility,is still dependent on manual reconfiguration(Over The Air or otherwise) and is not able toadapt in goal-driven scenarios withoutsubstantial operator/man-in-the-loopintervention.

In contrast, the CR is ideally able to performhigh-level cognition tasks autonomously. Takento extremes, the CR is able to propose andnegotiate protocols (Level 9) with minimal userintervention. This is akin to two humans whospeak different languages being able to

Table 1. Cognition Tasks (Mitola, 2000)

Cognition Capability Task CharacteristicsLevel

0 Pre-programmed Radio has no model-based reasoning capability

1 Goal-driven Goal-driven choice of frequency band, air interface,and protocol

2 Context Infers external communications context with Awareness minimum user involvement

3 Radio Aware Flexible reasoning about internal andnetwork architectures

4 Capable of Reasons over goals as a function of time, spacePlanning and context

5 Conducts Expresses arguments for plans / alternatives to users,Negotiations peers, and networks

6 Learns Fluents Autonomously determines the structure ofthe environment

7 Adapts Plans Autonomously alters plans as learned fluents change

8 Adapts Autonomously proposes and negotiates new Protocols protocols

communicate with each other (after a learningperiod) without outside help.

In truth, the CR need not attain Level 8 to pushradio communications to a new frontier. Withcontext awareness (Level 2), self-awareness(Level 3), planning capability (Level 4), andnegotiation capability (Level 5), the CR wouldalready have relieved radio operators of muchof the frequency preplanning and other tasksrequired for blue force communications.

Figure 2 shows the suite of functions whichmay be required in order to unleash the fullpotential of a cognitive radio. The radio mustsense the external environment, know its ownposition in this context (hence the need forpositioning), and be able to access multiplebands and networks.


21


POTENTIAL BENEFITS

Dynamic Spectrum Access

A major motivation for the development ofCR today is its promise of opportunistic anddynamic spectrum access. Today, licensedspectrum is a scarce and costly resource.However, studies conducted in the US indicatethat at any one time, in any one location, theactual spectrum usage is only 5-10%.

The CR senses spectrum usage by licensed usersand uses the unused spectrum, giving way tolicensed users when it detects that the licenseduser is transmitting a message. This applicationbrings tremendous benefits to the commercialtelecommunications industry, and is currentlya major driver in CR research.

In the military context, dynamic spectrum accessdoes away with much of the laboriousfrequency preplanning and coordinationrequired before deployment. Blue forceinteroperability is also enabled since twomutually interfering equipment in closeproximity would adapt to each other in sucha way that interference is reduced.

Increased Network Throughputand Reach

In a normal scenario, a sender who desires totransmit a message would need to sense if thedesired spectrum is currently occupied by asender. If it was, many current networkprotocols required the second sender to backoff. In contrast, CR offers the potential forboth senders to transmit simultaneously. Usingan algorithm similar to dirty paper coding, thesecond sender could encode his transmissionsuch in a way that the interference presentedby the second sender to the first receiver wastransparent to the first receiver. A prioriknowledge of the first senders transmissionby the second sender is required; this wouldbe obtained by the CR via its spectrumsensing methods.

Alternatively, an idle CR that detected atransmission from another node in its networkcould aid the transmission by transmitting acopy of the same message, thus increasing thereach and achievable data rate of thetransmission. Collaborative sending wouldneed to be balanced against each nodes ownneed for transmission. Network codingtechniques, with which a node in reach of both

Figure 2. Potential required functionalities of a cognitive radio5

-02_CognitiveRadio_p20_p22.FH11 4/10/07 9:28 PM Page 2

22

the sender and receiver nodes encodes themessages between them smartly to reduce thetotal number of messages sent between thetwo communicating nodes, could also beapplied in an adaptive / cognitive radio networkwith greater ease than a non-adaptivenetwork6.

In a non-collaborative scenario, the CR wouldincrease data rates and reach by sensing theinterference within its context, and adaptingto it when the need arises by tweaking itsown radio parameters e.g. transmittedpower, modulation scheme, or evenantenna directivity.

Interoperability

Due to the congested nature of theelectromagnetic spectrum, it is common formultiple emitters to share the same frequencyband. Thus, interference mitigation isunavoidable for blue force interoperability.CRs dynamic spectrum access functionalitygoes a long way towards mitigating much ofthe mutual interference. Smart transmissionencoding by secondary users of the spectrumas described above would also mitigate theinterference experienced by receivers. Inaddition to frequency access and encodingtechniques, CRs equipped with smartdirectional antennas and power control wouldalso focus their transmission only in thedirection of the intended receivers at anappropriate power level to minimise potentialinterference to other receivers.

Interoperability between different systems hasbeen a major concern for many military forces.As noted earlier, the US was motivated to driveSDR development simply because they haddifficulties communicating across their differentradio systems (Joint Program Executive Office,2006)! A high-level CR would mitigate thisproblem by negotiating and proposing newprotocols. Essentially, two Level 8 CRs wouldbe able to communicate by autonomouslyproposing new protocols between themselves.

On a less ambitious note, a Level 3 CR would

learn about the existing radios in its proximity

and depending on a priori knowledge, switch

to the appropriate protocol to communicate

with its desired receiver nodes. This does away

with the need to carry multiple radio

equipment or to implement a gateway for

each pair of radio types.

Communications Assurance

All users of wireless communicators such as

mobile phone users have had the experience

of a call being interrupted by noise or even

suddenly terminated as they moved from

location to location. A CR would sense the

changes in the external environment and adapt

based on the users requirements. For example,

in an especially noisy area, the CR could sacrifice

data rate for low bit error rates by changing

its modulation, or by simply changing the

frequency band. Over time, the CR would learn

the combination of parameters that works best

in each environment e.g. foliage vs. urban

terrains, and adapt by adopting the optimal

set of actions for each environment. Examples

of observed parameters (meters) and the

possible reaction parameters (knobs) are

shown in Table 2.

CHALLENGES

The challenges to realising a CR in the military

context are manifold, ranging from physical

considerations to policy development.

Software Challenges

CR has the ambitious goal of incorporating

and implementing the learning capacity and

flexibility of a human brain into a machine.

The tasks associated with this goal are

staggeringly complex. These range from basic

cognitive functions such as data collection from

the environment and user sensing, to medium-

level functions such as performing dynamic


23


resource allocation and radio parametric

adaptation in real time, and high-level

functions such as collaborating and negotiating

with ever expanding ad hoc networks and

applications. In addition, the military

operational scenario is unpredictable and ad

hoc, requiring communications any time and

anywhere, presenting a problem that is not

easily bounded or predictable. The artificial

intelligence skill set required to tie the CR

together is a key challenge that needs to be

overcome to realise this dream.

Radio Hardware Challenges

The CR needs a hardware platform, very

much like how a Microsoft Windows Operating

System needs a PC with which to influence its

position in the world. Although SDR technology

has progressed significantly since the inception

Layer Meters Knobs(Observable parameters) (Writable parameters)

MAC Frame error rate Source coding

Data rate Channel coding rate and type

Frame size and type

Interleaving details

Channel / slot / code allocation

Duplexing

Multiple access

Encryption

PHY Bit error rate Transmitter power

Received signal power Spreading type and code

Noise power Modulation type and order

Interference power Pulse shaping

Power consumption Symbol rate

Fading statistics Carrier frequency

Doppler spread Automatic gain control

Delay spread Antenna directivity

Angle of arrival

Other Computational power CPU frequency scaling

Battery life

Table 2. Examples of Meters and Knobs7

of the JTRS programme as well as other non-

US programmes, coupled with the maturity of

the common Software Communications

Architecture (SCA)8, a few finer details remainto be worked out. One of these finer detailsis the adoption of common security hardwarerequirements for the SCA-compliant SDRplatform.

The CRs hardware requirements are not trivial.For example, in order to dynamically accessspectrum without interfering with primaryusers, the CR would need -- (a) to have areceiver much more sensitive than that of theexpected receivers in the area; (b) to filter themultitude of incoming signals and decide whichare the valid signals from existing users; (c) tolocate the existing users; and (d) to adapt inreal time its own transmission behaviour interms of power, direction, frequency band,modulation, etc. The challenge for CR hardware


24

is to fit all of these requirements into a singlecost-effective and portable platform9.

Regulatory Challenges

In order for CR to interoperate on a large scale,rules of behaviour need to be set out andadhered to. For example, in the commercialcontext, in the case of dynamic spectrum access,the interests of licensed spectrum users needto be protected. Unfortunately, the currenttechnology is not sufficiently mature inproviding the necessary level of assurance toexisting primary users to reduce their resistanceagainst the deregulation of spectrum access.In the military context, etiquette would dictatethe behaviour of secondary users who wantto operate in the same spectrum and localitywhere other military primary users arealready operating. Such rules of behaviourcould help to mitigate the mutual interferenceeffects of emitters and receivers located onthe same platform10.

Policy development for CR is a necessary butdelicate and painstaking process involvingvarious stakeholders ranging from regulatorybodies to industry and R&D players whowould influence the policy trends withtheir technology.

Security Challenges

Over-the-air reconfiguration of a SDR offersan operational flexibility that is currentlyunheard of. However, a highly-evolved CR thatis able to negotiate with other radios andmodify its own protocols would be even morevulnerable to malicious software attacks.

Hence, there are security implications associatedwith the SDR/CR downloading, installing andusing software that could reconfigure the radiobehaviour such as frequency, power, andmodulation. Unauthorised modifications tothe SDR/CR, for example, could result in systemoverload, violation of power control profile,

unauthorised access to content, impersonationand spoofing behaviour. In the military context,the security implications cannot beunderestimated.

Security issues include the identification of theauthority to control the reconfiguration,protection of the reconfiguration signal,accuracy of the reconfiguration information,secure download of the reconfigurationsoftware, and conformance requirements ofthe hardware (Cook, 2004 and NTIA, 2005)11.

CONCLUSION

With the trend towards network-centricityboth in the commercial and military arenas,the use of wireless communicationstechnologies and devices is becoming moreprevalent everyday. The solution to meetingthis insatiable appetite for connectivity, in theface of limited frequency spectrum and rigidregulatory policy rules, while not increasingcosts (operator or man-in-the-loop), lies in thespirit of CR.

This dream is not beyond reach. The DefenseAdvanced Research Projects Agencys NextGeneration (XG) programme is developing thetechnology and system concepts to dynamicallyaccess the radio spectrum. The first in a seriesof field demonstrations of XG integrated testswas recently conducted and showed that XGcould operate without interfering with existingradios in the area12.


25


6. J. Mitola III, Cognitive Radio -- An IntegratedAgent Architecture for Software Defined Radio,Doctor of Technology Dissertation, RoyalInstitute of Technology (KTH), 8 May 2000

7. National Telecommunications andInformation Administration Comments ETDocket No. 03-108

8. J. Reed, C.W. Bostian, Understanding theIssues in Software Defined Cognitive Radio,Mobile and Portable Radio Research Group,Virginia Polytechnic Institute and StateUniversity, U.S.A

9. T.W. Rondeau, B. Le, D. Maldonado, D.Scaperoth, C.W. Bostian, Cognitive RadioFormulation and Implementation, CognitiveRadio Oriented Wireless Networks andCommunications (CROWNCOM) 2006

10. Joint Program Executive Office, JTRSOverview Brief, 2006

REFERENCES

1. P. Cook, Wireless Software DownloadSecurity, SDR Forum Document No. SDRF-04-I-0069-V0.00, June 2004

2. N. Devroye, P. Mitran, V. Tarokh, AchievableRates in Cognitive Radio Channels, IEEETransactions on Information Theory, vol. 52,No. 5, May 2006

3. S. Katti, H. Rahul, W. Hu, D. Katabi,M. Mdard, J. Crowcroft, XORs in the Air:Practical Wireless Network Coding, SIGCOMM2006

4. M. McHenry, Presentation to FCC Workshopon Cognitive Radios, 2003

5. J. Mitola III, Cognitive Radio for FlexibleMobile Multimedia Communications, IEEEMobile Multimedia Conference, 1999


26

packets. X1 can then recover X2s data by

performing an XOR of its own data with the

data from the relay node. The same applies to

X2. The total number of transmissions required

is reduced to three. Experimental results

indicate that the network throughput gain

varies from a few percent to several times

depending on traffic patterns, protocol,

congestion levels.

7. Courtesy of Centre for Wireless Technology,Virginia Polytechnic Institute and StateUniversity.

8. The latest released SCA version is SCA v2.2.2.

Note: SCA v2.2 is most commonly adopted

by industry players.

9. The authors note that the signal detection

threshold is higher than the receiver

sensitivity in order to reduce false alarms.

Thus, the CR may not detect an existing

signal that is very weak even if the signal

level is above its receiver sensitivity and

transmit, causing an increase in interference

to the existing user.

10. Today, the conflicts between emitter(s) and

receiver(s) on the same platform are resolved

via sending blanking signals to the affected

Systems when the emitter wishes to transmit.

11. Please see Ref [1] & [7].

12. Please refer to Shared Spectrum

C o m p a n y s 1 8 S e p t e m b e r 2 0 0 6

p r e s s r e l e a s e a t :

http://www.sharedspectrum.com/content/pre

ss/XG_Demo_News_Release_060918.pdf

1. This is the Latin rendition of Descartesoriginal statement Je pense, donc je suisin 1637.

2. A SDR is a radio whose channel modulationand demodulation waveforms are definedin software. As the purpose of a SDR is toallow re-programmability simply via loadingnew software, the SDR performs much ofits signal processing on re-configurableelectronics such as general purposeprocessors.

3. Many cognitive radio researchersperceive the Cognit ive Radio as abiological organism due to its ability to learnand adapt to external stimuli.

4. Disruptive technology is a term coined byHarvard Business School professor ClaytonM. Christensen to describe a new technologythat unexpectedly displaces an establishedtechnology.

5. Adapted from Joseph Mitolas Doctor ofTechnology Dissertation, 2000.

6. MIT CSAIL and University of Cambridge(Ref [3]) have worked on a new forwardingarchitecture for wireless mesh networkscalled COPE. COPEs advantage can be seen inthe simple scenario of two nodes, X1 andX2, communicating across an intermediaterelay node. In most network architectures,a total of four transmissions would be required

for X1 and X2 to communicate. Under COPE,

the relay node stores the data from X1 and X2

and transmits an XOR of these two data

ENDNOTES


27

BIOGRAPHY


Liew Hui Ming is Programme Manager (Directorate of R&D). He is responsible

for developing and charting out the Long Term Defence Technology Plan for

Advanced Communications as part of the IKC2 R&D portfolio. Hui Ming led

the team which received the DSTA Excellence Award in 2003 for the Tactical

ONE Network Experimentation. He was also part of the team which received

the DSTA Team Excellence Award in 2006 for the Battlefield Instrumentation

System. Hui Ming holds a Bachelor of Engineering (Information Systems

Engineering) (Honours) from the Imperial College, London, where he also

received the Imperial College Psion Prize for Best Final Year Project in Adaptive

Filtering. He also holds a Master of Science (Electrical Engineering) from the

National University of Singapore.

Lee Kwee Geak is Project Manager (Directorate of R&D). Her work focuses onnew Wireless Communications R&D initiatives such as the Software DefinedRadio. She was previously in the Sensor Systems Division, and worked onSurveillance Radars. Kwee Geak graduated with a Bachelor of Science (ElectricalEngineering) summa cum laude with honours from Cornell University, US,under the Defence Technology Training Award. She also holds a Master ofEngineering (Electrical) from Cornell University. Kwee Geak received theHarvard Prize Book in Economics in 1997.

-03_JEWEL_p30.FH11 4/10/07 9:29 PM Page 1

30


INTRODUCTION

As a small nation, Singapore has limited humanresources, land and airspace. The strategic useof Modelling and Simulation (M&S) to help usovercome these constraints is therefore crucial.We thus formulated our simulation masterplan, called the Vision for SAF Simulations(VSS), back in the mid-1990s. The JointModelling and Simulation Environment forWargaming and Experimentation Labs (JEWEL)was conceived as the simulation environmentin support of VSS. This enterprise-wideapproach to simulation is analogous to whatis happening in the business and commandand control (C2) worlds.

Designed with reusability and interoperabilityas its primary precepts, JEWEL would be anopen software environment that allows theincorporation of new technologies andstandards from governmental, commercial andR&D bodies. It would be a launching platformfrom which new application needs can besatisfied accurately and quickly. To maintainopenness and as a result future-proof JEWEL,DSTA believes that substantial attention mustbe devoted to its information architecture,both in terms of representation as well ascontent, as demonstrated in our adoption ofHigh Level Architecture (HLA) and ExtensibleMarkup Language (XML), among otherstandards. JEWEL would support the SingaporeArmed Forces (SAF) in training, analysis,experimentation and acquisition.

JEWEL

Motivation

The primary motivation behind JEWEL is theanticipated increase in demand for M&S in theSAF to support experimentation, training andoperations. We can no longer afford to relyon the traditional approach of developingstovepiped simulation systems for eachapplication. A more radical approach thatinvolves the development of a common,interoperable and shared M&S environmentwill be required, with these long-term benefits:

Reduced Cost, Shorter Time-to-Deploy,Reduced Risk. Through the reuse of commoncomponents and models, the cost to developnew M&S applications would be greatlyreduced, as we only need to develop the deltasto meet the specific user requirements. This,in turn, translates to shorter time required todeploy the systems. Project risks would also bereduced, as we are reusing tested componentsfrom similar projects.

Meeting User Requirements On Demand.With composability built into the environment,the systems deployed will be more flexible andre-conf igurable to meet d i fferentexperimentation requirements, on demand.

G r e a t e r C o m m a n d , C o n t r o l ,Communications, Computers and Intelligence-Simulation and Simulation-SimulationInteroperability. The focus on shared conceptualand data models, common interoperabilitystandards and components will enhanceinteroperability among simulations, and withthe C4I systems.

Improved Consistency. Standardisationfacilitates the verification and validation ofmodels. Reuse of these validated models willenhance the consistency of outcomes acrossour M&S systems, which is key if M&S were toaid decision-making.

Extended Systems Shelf Life. The adoptionof an enterprise architectural approach ratherthan the stovepipe system approach will lendgreater manageability to continuously upgradeour systems in order to keep up with newtechnology and standards, thereby extendingthe shelf life of these systems.

Anatomy

JEWEL is a collection of data and interfacespecification standards, frameworks and tools,and composable models and databases, asshown in Table 1. The rows of the table areexplained as follows:

Repositories of Composable Models andDatabases. This layer refers to the physical set

-03_JEWEL_final.FH11 4/4/07 1:34 PM Page 4

Table 1. Anatomy of JEWEL

31

of reusable models and databases that aredeveloped or acquired to perform specificsimulation functions.

Frameworks and Tools. This layer consists ofthe technical frameworks that provide the glueessential for the different simulation tools tointegrate seamlessly with one another. Someof these tools are available commercially off-the-shelf while others are custom-developedfor the JEWEL environment.

Data and Interface Specification Standards.In order to enhance the level of interoperabilityamong simulation systems, and betweensimulation and C4I systems, it is important thatthese systems adopt consistent data andinterface specification standards. This wouldminimise the problems associated with the

interpretation and mapping of data as a resultof representation differences.

Across Table 1, we account for all activitiesduring the M&S lifecycle. The columns areexplained as follows:

Modelling. Model development poses greatchallenges to system developers, as it requiresboth the technical expertise of programmersand the domain knowledge of subject matterexperts. Maintaining models as technologychanges is also an issue. The direction istherefore towards a formal developmentapproach whereby models are captured at theconceptual level, independent of theunderlying platforms and technologies. Thegeneration of platform-specific codes shouldbe automated as far as possible.

Repositories ofComposableModels andDatabases

Frameworkand Tools

Data andInterfaceSpecificationsStandards

Models (and Specifications) RepositorySim-C4I Interface Maps Repository

Frameworks and Tools RepositoryReal-time Platform Reference Federation Object ModelsBattlespace Repository

Integrated On-Demand SystemModel Synthetic ConfigurationDevelopment Battlespace Editor Environment ComposerTerrainDevelopmentTools

Data AnalysisConsoleAfter ActionReview ToolsDataAnalysisTools

ScenarioEditorDistributedSimulationEngineComputerGeneratedForcesFrameworkInjects FrameworkAssortmentof GraphicalUser

Models SpecificationsSim-C4I Interface Maps Repository

System Configuration Repository

Simulation State Repository

A c t i v i t i e s

Tools Interface SpecificationsHigh Level Architecture Federation Object Models SpecificationBattlespace Specification

System Configuration Specification

Simulation State Specification

AnalysisSimulationDeploymentand System

ConfigurationModelling Compositionand Integration


32

Composition and Integration. Battlespacecomposition adopts the lego brick paradigmfor M&S system development. It stems fromthe desire to re-configure and reuse systemsthrough plug and play. It is often a statedsystem objective among the more recent M&Sdevelopments such as OneSAF and is a resultof the advancements made in component-based technologies.

Deployment and System Configuration.M&S systems today do not normally operatein a standalone manner. They are oftenembedded within C4I systems, and operatewithin a federation comprising other simulationsystems. HLA is the current technology forintegration among simulation systems. Thetechnology is still being enhanced to addresssome of its current limitations. Integrationbetween M&S and C4I systems is traditionallyachieved through the use of dedicatedgateways. Increasingly, web technologies arebeing considered as the enabler for greaterSimulation-Operations integration.

Simulation. This is the focal point of allsimulation environments as it is where theactual simulation is being executed. All theother stages exist to support the runtimesimulation. The key issues that characterise thisstage are mainly performance-related, such asthe scalability of the specific runtime simulationarchitecture and the level of interactivity. Asimulation engine is usually at the heart of theruntime architecture, which in turn is supportedby other simulation components such asbehavioural or Computer Generated Forces(CGF) engines.

Analysis. Analyses are conducted in orderto maximise the values of the simulation runs.Standard after-action-review features includethe ability to perform record and playback.More advanced capabilities would include theability to use commercial off-the-shelf toolsfor statistical analysis as well as data miningof data captured during simulation runs.

We devote the entire next section to theDistributed Simulation Engine (DSE), as it isthe core component of JEWEL, and plays apivotal role in our endeavour to achieve reuse

and interoperability. DSE also embodies highperforming design and implementationdecisions, in order to satisfy the disparate needsof various M&S communities. These includeexperimentation and training.

DISTRIBUTEDSIMULATION ENGINE

Introduction

DSE is a software framework that providesthe following:

A set of base classes that represents thecommon concepts. By default, these classesdefine a set of Application ProgrammingInterfaces (API) that different components ofapplications may use to interact with oneother (see later sections). These base classescan be used to define application specificconcept types.

A simulation kernel that keeps track of timeand orchestrates the execution of allapplication components.

Communication capability that handlesnetworking issues between processes builton DSE.

Distributed database services for storingruntime instances of all concepts.

Avenues for attaching commonly usedrepositories or libraries of functions. Examplesinclude the assets definition database, scenarioloader/writer, simulation state record/playbackmodule and environment database, etc.

Key Concepts Defined

We have defined four main concepts thatrepresent the software equivalent of thereal-life battlefield:

Entity. This refers to any object that has thecapability to act and react, which can includebrigades, squadrons, tanks, soldiers, task forces,missiles, radar, etc. An entity contains attributesthat define its characteristics, such as the


33


amount of fuel, number of missiles, travelvelocity, level of casualty, physical shape andsize, etc.

Model. Whereas entities represent theobjects that can act and react, models definehow they perform such action and reaction.For instance, an action of one entity is to firea missile, and the reaction of another entity isperhaps to compute the damage sustained.This concept includes the followingsub-concepts:

- Behavioural Model. Analogous to thehuman brain, this defines the decision-making processes of each entity. Forinstance, an aircraft performing a civil airpatrol may have to deliberate betweenengaging an approaching enemy fighteror not.

- One-time Action Model. Action modelsimplement the exact physical level activitiesthat are performed by the entities.A distinction is made between activitiesthat must be performed throughout thelifetime of the entities and those otherwise.One-time Action Model refers to the latter.

- Repetitive Action Model. This definesactivities that must be performedthroughout the lifetime of entities.

In addition to representing entitiesbehaviour, the model concept also extends

to modules that display a Graphical UserInterface that interacts with users andmodules that handle the networking aspectsof DSE.

Models in DSE are classified into twocategories: Global or Entity-based. Entity-based models contain codes specific toindividual entities (e.g. models thatimplement the action of an entity),whereas Global models maintain visibilityof all entities (e.g. User Interface modelsthat provide avenues for users tocontrol the simulation).

Event. Events represent the happeningsin the battlefield. Events are created when

something that may be of significance to otherentities happens. The models that performsensor operations usually receive these events.

Command. Commands represent instructionsor orders given to the entities. For example,an order to attack an enemy position willinclude the identification information of thatposition, contained in a command object. Thisconcept includes the following sub-concepts:

- Behavioural Command. Commandstargeted for Behavioural Models.

- Action Command. Commands targetedfor Action Models.

Command is also a convenient mechanismfor status reporting. That is, the recipient ofa command can use that same command toconvey the status of task completion backto the command issuer.

DSE was written in C++, and was designedunder the premise that all simulations can, andshould be built by simply specialising thebase classes (class inheritance in C++nomenclature) of these concepts. Thus, creatinga new model class by inheriting from the Modelbase class (defined in DSE) and insertingadditional attributes and logic (by means ofoverriding virtual functions) that are peculiarto how an F16 aircraft manoeuvres wouldcreate an F16 aircraft motion-dynamic model.

Design Considerations forScalability

Scalability Defined. Scalability refers to theability of a simulation system to support largerscenarios. Larger scenarios encompass anincrease in one or both of the following: thenumber of simulated objects and the size ofthe battlefield. In reality, no software scalesinfinitely, due to the limits imposed by theOperating Systems (OS), processing hardwareand network infrastructure.

Objects Count. DSE was designed to representup to 263 objects during each simulation,although the processor speed and networkbandwidth will further limit the number ofsupportable models or processes.


34

Process Distribution. DSE enables the splittingof a simulation application into processes, eachsimulating different objects or parts of objects.Such splits can be attained without adding onto the load of the inter-simulation network(typically a High Level Architecture-RuntimeInfrastructure network), because DSEimplements its own intra-simulation network.The provision of an intra-simulation networkhas these additional benefits:

Allows more intimate relations betweenmodels within the same application, therebyincreasing modelling efficiency.

Able to scale without relying on HLA. Theintra-simulation network can also be enhancedindependently, should new and betternetworking technologies surface.

Enable scalability without affecting HLA,especially during scenario creation. HLAsupports runtime coordination amongapplications well, but does not contribute toconvenience during scenario creation.

Design Considerations forExtensibility

Extensibility Defined. Extensibility refers tothe ability of a simulation application tosupport new kinds of objects. In a sense allapplications are extensible, just that extendingsome requires greater effort than the rest.

General Techniques. DSE is an Object-Orientedsoftware that is highly modularised, with well-defined interfaces that connect to allapplication components (the concepts, definedearlier). Applications built on DSE must packagetheir models into Dynamic Loadable Libraries(DLL), so that they can be loaded and unloadeddynamically during runtime. Another benefitof using DLLs is that models can be addedwithout requiring re-compilation or re-linkingof any software component.

Extensible Parts. All concepts defined earlierare extensible with DSE. All extensions toapplications built on DSE can be performedwithout resulting in code compilation, exceptnew models, in which case coding and

compilation is necessary. This effectively meansthat non-technical end-users may extend asimulation by adding new entity types, togetherwith their characteristics, without looking ata single line of code. Apart from models,entities, events and commands, databases forscenario and environment representation canalso be plugged into DSE easily, so long as thepre-defined APIs are adhered to.

Design Considerations forEfficiency

Efficiency Defined. Efficiency is the measureof the throughput of a simulation against time.Basically, the more objects that can besimulated within a fixed time, the moreefficient the software is.

General Techniques. Multithreading andparallel execution by process distribution aretwo general performance-enhancingtechniques employed by DSE.

Localised Memory Allocation/Re-allocation.The time consumed by the operating systemduring new memory allocation is notpredictable, and the delta can range from justa few microseconds to a few milliseconds. Toovercome this potential performance letdown,most simulation software reserve a large poolof memory at start up and implement theirown localised memory allocation or re-allocation routines. DSE goes one step further,to also reuse memory lots previously allocatedto data of the same type. This further shortensthe time needed to allocate or re-allocatememory, and paves the way for more efficientlogging and networking functions.

Runtime Model. The two well-establishedruntime models adopted by simulation systemsare time-stepped and event-driven models.Conventionally, applications that require ahigh level of user-interactivity adopt the time-stepped model, while non-real-time and non-interactive applications take advantage of theevent-driven model for efficiency reasons.

DSE is time-stepped driven with an adjustableframe size and rate. A new simulation state iscomputed in every frame. If the frame size is


35


x ms and rate is r, DSE will invoke all applicationmodels at x ms interval and the simulationclock will advance by x*r ms, as illustratedin Figure 1.

Figure 2 depicts the perfect scenario, in whichthe computation time consumed to calculateeach state falls within x milliseconds. For DSEto handle exceptional scenarios, three timingmodes were implemented:

Race Mode. Simulation states are computedone right after another. This is useful forapplications that do not require userinteraction, and only the final state isof importance.

Disregard Real Time Mode. If the time tocompute state n exceeds the frame interval,computation of state n+1 commencesimmediately after state n is computed.However, if the time to compute state n isshorter than the frame interval, the kernel willwait until the frame interval elapses completely,before computing the next state. In both cases,the simulation time advances by the sameamount the frame interval. Applicationsmay assume that each computationcomputes the next state of objects for fixedtime advancement.

Follow Real Time Mode. Advancement ofsimulation time is tied to real time, at thespecified rate. Computation of one state maybe skipped if the previous state requires morethan the frame size to compute. Applicationsmust be built to compute states of objects forvariable time advancement.

An enhancement to thestandard time-steppedmodel is to implement ahybrid runtime model. Thiswill be elaborated next.

Execution Scheduling Policy.It is important to note thatthe battlefield is bestrepresented if all modelsexecute in parallel andcontinuously. However, this

is not possible, as it would require one CentralProcessing Unit for each model. As such,some scheme of scheduling the executionof the models within each processis necessary.

On top of the basic runtime model, DSEimplements an efficient scheduling algorithmthat supports advanced scheduling options,as follows:

Time Schedule Order. Much like event-drivensimulation, this feature will enable models toschedule their own execution time. However,unlike event-driven simulation, execution iscarried out only at the nearest later frametime, rather than immediately. Entities, modelsand in fact any application function can bescheduled at a time decided by the application.

Variable Frequency. Useful in handlingmodels of differing updating resolution. Forinstance, the position of a tank on the groundcan be updated at one second interval, whilean aircraft requires at least 100ms.

The advanced scheduling algorithm is able toimprove the overall efficiency of simulationsbecause of its ability to spread out theinvocation of low frequency entities andmodels. For instance, in a scenario with twoaircraft and 10 tanks, only one aircraft andtwo tanks need to be invoked for computationin each frame if the update resolution ofaircraft is 200ms, that of each tank is 1000ms,and the simulation frame interval is 100ms.

sr sr+x sr+2x sr+3x sr+4x sr+5x sr+6x

ss ss+x*r ss+2x*r ss+3x*r ss+4x*r ss+5x*r ss+6x*r

Realtimeline

Simulationtimeline

Legend: sr - start time (real)ss - start time (simulation)x - frame sizer - frame rate

Figure 1. Effects of Frame Size and Rate


36

Design Considerations forReliability

Reliability Defined. Reliability refers first tohow frequent downtimes occur for anapplication. Applications with high reliabilityexperience a lower frequency of downtimes.Ideally, there should be no downtime at all.As DSE is an extensible software framework(i.e. it will execute together with non-nativecodes), it is unrealistic to aim for the highestlevel of reliability (i.e. no downtime). Therefore,DSE provides a mechanism to ensure quickrecovery from such downtimes.

Data Replication. There are generally four waysfor data to be shared among the processes ofdistributed applications and each has its prosand cons, as described below. DSE applies thesecond method:

Centralised Database. This is the traditionalapproach, in which the entire simulationdatabase is hosted on one server. All processesconnect remotely to access the database.Although slow in access speed, data consistencyis maintained with ease. Adopting mirroringtechnology will remove the single pointof failure.

Replicated Database. The database isreplicated across all processes. While utility ofnetwork goes up and maintenance of dataconsistency is tough, this method incurs lowdata access latency and allows flexibility interms of data and scenario size.

Shared Memory. Using a combination ofhardware and software, this solution allowsprocesses across multiple machines to sharethe same memory spaces. One major drawbackof this method is its limited address space andthe fact that the failure of any singlecomponent will result in complete systemfailure, although the lowest data access latencycan be expected.

Scalable Parallel / Message PassingInterface. Housing all processes of oneapplication in one Symmetric Multiprocessingmachine that also hosts the database, this

method often produces the highestperformance for special ly designedapplications. The drawback is the proprietarystatus of both hardw

dsta horizon 2007

Documents

pm page

joint warfare

dsta horizons2007

thisissue of dsta horizons

radio systems

smart radio

joint battle system

defence development