An encapsulation for reasoning, learning, knowledge representation, and reconfiguration cognitive radio elements

(1)

An

Encapsulation

for

Reasoning,

Learning, Knowledge

Representation, and

Reconfiguration

Cognltive Radio Elements

Keith E. Nolan

*

Paul

Sutton

Linda E. Doyle

CTVR

Trinity College Dublin

Trinity

College Dublin

Trinity

College Dublin

[email protected]

ledoylegtcd.ie

Abstract

behavior, respond to the time-varying nature of wireless channel and user

activity

and learn from previous experi-State and contextual awareness, reasoning and con- ences. Considered in

isolation,

the foundations of each of clusions

formation,

andameans of directing

application,

thecore

observe,

orient, reactand learn stages of the cog-structural andparameter-level radio reconfiguration are nition

cycle

first described

by

Mitola L]andmorerecently key elements ofa cognitive radio. This paper describes

by Haykin

[2],

are not new concepts. However, it is the a cognitive radio design capable of scaling between the innovativeapplicationofacombination ofthesetechniques

two extremes of minimal cognitive capabilities and com- inacognitivewireless communicationscontextthat is

inno-plex highly-evolved cognitive radio

abilities,

which is be- vative. Observation informationcanbe derived from inter-ing adopted for real tests using licensed cognitive radio nal radio and system

activity (including

available resources, testspectrum. A memory element storesstate,

sensor,

ob- radio

capabilities

and spectrum

activity

detectedat the

re-jectives,

actions and conclusions information and the rel-

ceiver),

and external sources

(including

external environ-evanceofthis information canbe varied in order to iden- mental sensors,

policies,

network-level

information).

tify orignore common traits or occurrences. The

decision-makingandconclusionsformationabilitiesofthis cognitive An

implementation

of a

cognitive

enti

requires

a radio design can use (or choose to ignore using the vari-

highly-reconfigurable

core, which can

change

and evolve able

weightingfacility)

external

information relating

to the

according

tothe

orient,

reactand learnstagesinthe

cogni-network and

etiquettes

in

conjunction

with the memoryee-

tive

cycle. Popular approachestaken

in

relation

tohow

and

ment. Asetofactionsformulated bythe reasoning and con-

why changes

arenecessaryarebasedon

game-theoretic

[5],

clusions

formation

stages direct the radio

reconfiguration.

genetic

algorithmic

[6], Fuzzy

Logic

[8] and

artificial

This

design

is

implemented

using

a

General-Purpose

Pro- neural-network

principles

[7].

These

techniques

are used cessor

(GPP)

platform

as it

currently offers

the very

high

to

find.

an

optimum (or near-optimum)

solutionto a

paic-level

ofreconfigurability requiredfor

verymalleable

cogni-

ular

wireless communications

problembut

require

a recon-tiveradio

design.

figurable

radio

in orderto

implement

the desired changes

andanalysis theimplications of this change. Itis feasible that fully-engaged cognitive abilities are always required

1. Introduction

depending

on the

complexity

of,

and

challenges presented

by particularscenarios. Therefore,the

ability

tochange be-This section introduces the idea of

cognition

and identi- tweenminimal and

complex

cognitive

behaviorcan

poten-fies thecore

requirements

fora

cognitive

radio.

tially

reduce power

consumption

and increase the

operating

Cognition

in

signal-processing

andsystemcontrolterms lifetime of the device. is the

ability

to develop contextual and, environmental

awareness aidingthe development of an optimal solution Section 2 describes thereconfigurable

core,

which is the for aparticular problem, recognise developing patterns of keyenablingfeature for thiscognitive wrapper, Section 3 is a description of the core

elemenats

of the coglnitivewrapper

*This material is based upon work supported by

Scienlce

Fourndatioun inldn th awrns prcsig

le.nn

atoman IrelanLd unlder Gralnt No. 03/CE3/I405 aspar of the Celntre for

Telecom-munlications Vallue-ChainlResearch (CTVR) atTrinlityColllegeDubllin,tre- tiation stages. InSectionL4presenLts solme conclusionks frolm

(2)

2,

Reconfigurable Radio

2.1.

Reconfiguration

Drivers

Preparing

for the

possibility

of

change

within a

cogni-This section briefly describesthe term

reconfigurable

ra-

tive

radio does

not

imply thatcontinuous change is

required,

dio and one instance of an actual system that is used as the during the operating

lifetime

of the radio. It is conceivable basis for the cognitive wrapper described in this paper. that a static architecture is sufficient in some cases. Recon-The term reconfigurable radio is used in this paper to figuration

activity

is triggered when the cognition engine

describeaheteromorphicradiosignal-processing chain im- determines that an observed event(s) or states necessitate an plementedin software, connected to a minimal hardware RF application, component orparameter change. It is

necessary

front-end that may itself be reconfigurable through physical to drive these reconfiguration processes in an intelligent change or under software control. Areference design and, manner that will result in the implementation of a desired implementation of a reconfigurable radio used as the basis preset feasible solution or a solution developed. by the rea-of thereasoning wrapper design in this paper is calledIm- soning and conclusionsformation engine. Reconfiguration plementing Radio In Software (IRIS)

[9].

This system uses drivers are not limited tointernaldeviceevents/observations eXtensibleMarkup Language (XML)

[3]

todescribe aradio however, and can account for radio,

network,

regulatory in termsof a signal-processing chain of elements called Ra- and physical environment changes, and application, busi-dio Components. Examples of existing Radio Components ness and social context changes.

include modulators, demodulators, access schemes, filters,

signal

conversion,

source and sink elements. These Radio

Observations

Components can either be sourced.from a local inventory

of available

Components

(created

by

the

designers)

orfrom Awareness one or moreremotely-locatedinventoriesusingawired link.

Reconfigurable

Processing Each of these elements hasa commonarchitectural frame- Core & workfacilitating rapid developmentandstraightforwardin-

Reasoning

ternalcreation, executionand tear-down processes.

Action

Set

The IRIS system caters for a

hierarchy

of

possible

recon-figurationtasks called action sets. These action sets are de- Figure 1. Reconfigurable core showing

in-veloped, bythe observationsreporting, awareness process- puts (set of actions) and outputs (device

ing and reasoning engine loop as depicted by Fig. 1. Ap- state, capabilities and spectrum observa-plication reconfiguration allows the replacement of an en-

tions)

tire signal-chain with another desired signal chain in or-der to change the active application. Component

recon-figuration enables one or more signal-chain processing

el-ements(Radio Components) to be removed/replaced/added 3.

Cognitive

Wrapper

atwill. This reconfiguration can duringrun-time in

addi-tiontothe trivial static-case reconfigurationscenario. Dy- In this section, the

primar

contribution of this paper namicparameter-level reconfigurationis alsopossible and is presented in more detail. This is a realisable cognitive all relevant parameters used in each Radio Componentcan wrapper with

scalable-'intelligence'

and

designer-specified

bechangedondemand. Thesereconfiguration possibilities learningand

reasoning

algorithmcapabilities. This section allow the radiocore tobe molded into any formaccordingto describes the core entities

comprising

this cognitive wrap-the instructions ofahigher-level

entity

(cognitive wrapper), per design, where the key fundamentals of the design are which inturnispossiblyin responseone or morereconfig- shown in Fig. 2.

urationtriggersordrivers. The cognitive wrapper described in this paper encapsu-The higher-level IRIS

entity

governing

change within lates areconfigurable core, which in this case is the IRIS the reconfigurable core is called Control Logic [9]. system. Features of this wrapper, as illustrated in Fig. 2 This is a software mechanism that implements the re- include the observation, awareness and knowledge

repre-place/add/removeRadio Components and controls the cas- sentation mechanisms, a variable-length memory delay-cade of reconfiguration required when parameters are line used to store current and historical knowledge sets,

changed within one or more Radio Components that may which can also be used,to

identify

(or disregard) common impact on other

fRadio

Components

further

alonlg

the signal-

traits/characteristics.

The

reasoninag

engilnegenerates the re-chain. This Control Logic has beenexpanded.to cater for a configuration tasks andd.irectsthese changes in the recon-reasoning

engine, :me:mory delLay-lLine

and external input in- figurable core. This diagram also

ilUlustrates

thatconstraints

(3)

of the system capabilities and regulatory policies can also ing components,

data-type

descriptions, available process-have a direct influence on the reasoning and conclusions- ing power, RF front-end capabilities, networking capabili-formation processes. ties and, fixed, mobile, nomadic

mobility

status information. Reasoning tasks includedeveloping the sequence of ap- Extra sensing information can be obtained, from envi-plication, structural andparameter changes

(action

sets) or ronmental, spatial and biometric sensors including temper-deciding that no reconfiguration is

necessary.

Weconsider a ature, pressure, air and water quality, shock and vibration full-featured highly-involved, cognitive radio device for the information. Spatial awareness is not limited to geographi-following descriptions of the reasoning wrapper capabili- cal location but include trajectories, altitude and device-tilt

ties. information. Awareness of the time-value of this

informa-Implementation of a cognitive system requires tion is acritical factor in thecognitivecontrol mechanisms. awareness-formation, reasoning and learning, and, Available spectrum may haveafinite usagewindow, reac-conclusions-development capabilities. A cognitive tion to asudden shock experienced by thecognitive radio radio therefore requires a means of observing the envi- mayrequireimmediatecountermeasuresandadevice faced ronmental, social, user, spectrum and policy landscapes withadwindlingenergysupplymayhavetoinitiate grace-as described in Section 2.1, memorising (or choosing not ful

degradation

or

backup

measures before the remaining

to remember) previous events, actions and consequences, energy is depleted fully. Instead of the power-inefficient decision-making and conclusions-formation. The

ability

casewhere allpossible sensing sources are activated,atall tomould thereconfigurable core byexecuting actions that

times,

the

cognitive

radiomust be capable offocusing its direct theoperation and structure of this core is also a high resources ofsensingsourcesdeemedimportantatany

par-priority

objective. For maximum system flexibility, the ticular time and deactivatesensingsourcesconsidered irrel-radio device should have the ability to scale the influence evant.

ofthecognition capabilitiesbetween the two extremes of a

highly-involved cognitive radio to a basic device with no 3.1.

Knowledge

Representation

cognition capabilities.

Aspects

of the

radio,

including

current

and

obser-KnowledgeRepresentationDelay-Line i m

Memory{Tasks:Actions:Outcomes:

Conclusions} vationare

knowledge

sourcesusedaspartof theradio

cog-Obevton O1 2 3 N

Observationims

O X z s ...*s N nition processes. Itis the relevance of this

infoirmation

in l

fStotTeem

Long

Term aparticularcontext, instance orperiod of time, or scenario

Reconfifgurabl Vaibl that influences the value ofthis

knowledge

however. An

Core

.

oiguVabie

ability

to store the sequence of actions taken and

measur-0 able consequences of these actions is therefore avaluable

li

Decision-Making, Learning, asset. Information derivedfrom some source entity often

Actionl ket Conclusionls Formation I

Actionet ConclusionsFo n Variable hasastrict

description

syntax.

Inorderto

interpret

this

in-I____

0 e

Weighting:

formation

correctly

therefore, devices must conform to a

System egulatory common

syntactical

convention.XML

(eXtensible Markup

Constraints Policy

Language)

for

example,

is a

portable

method of

represent-Corintst§

ing information,which canbe parsed by softwareprocesses and is presented in a

human-interpretable

form

[3].

Web

Figure 2. Reasoning wrapper overview

illus-

Ontology Language

(OWL)

isamethod of

representing

in-trating the

knowledge

representation

delay-

formation that doesnot

necessarily

have to be

presented

ina

line, reasoning and

learning

engine, con- human-readable form but this information is

essentially

de-straints,

and

reconfigurable

coreentity rived from an

English language description

of the scenario ortask

[4].

OWL offers a meansof

specifying

the seman-tics ofascenario,whichcanbeconveyedand translatedby

A

highly-evolved cognitive radio

can employ contextual platforms with differentsyntaxconventions.

reasoning

to

help determine the best

course of

action

to

The ability

to store,

order,

extract and reuse

informa-take. Interpretationof selectedinternalandexternal physi- tionrelatingto currentand historical state,actions,

conclu-cal, spatial, environmental, political andobjectivesis there- sions, objectives in astructured format facilitates

applica-fore necessa to

develop

and maintain contextual aware- tion of this information in the cognitive decision-making

(4)

forward-planning and anticipative action ofthe cognitive ra- ficial neural networks, Bayesian or Fuzzy system logic im-dio. Emerging problems can be decomposed into a set of plementation approach. This is achieved using the Control problems with less

complexity.

Solutions to these nested Logic interface that provides the means by which, external challenges may already exist within the stored knowledge processes can attach to, and direct the reconfigurable core. setsthus potentially reducing the overall solution-formation This stage can also be de-activated using this Control Logic

time. interface if a minimal-cognition or non-cognitive device

op-The memory delay-line shown in Fig. 2 is the means used eration is required.

to store currentand historical knowledge sets for the cogni- The

feasibility

of a decision making, learning and con-tive radio system described in this paper. Amethod used to clusions

formation

approach is

dependent on

the time re-representshort and long-term knowledge, which formspart quired to present viable solutions and the implementation oftheinput and ultimately influences the reasoning wrapper

complexity

associated with each approach. The

presenta-outputs

and desired actions. Analogous to a finite-length fi'- tion of a solution approaching

optimality

within the time terdelay-line which stores current and historical knowledge constraints allowed has apotentially greater value than an sets. Information from all stored

memory

sets is available optimal solution that isproduced too late

i.e.

after

the im-for useby the cognitive engine. plementation deadline. Complexity and the processing bur-The relevance of certain aspects of each knowledge set den can be reduced by

implementing

some features of a stored in the memory delay-linemay not be constant. A chosen approach. It isconceivable that significant gains us-memory-merging

capability

offers some interesting possi- ing dynamic spectrum access techniques can be achieved bilities. For some

scenarios,

identification of common without the

full

weight of a

maximal-complexity

cognitive traits, actions or consequences of previous radio reconfig- engine. The platform presented in this paper offers the uration and observed events may be more important than

ability

to investigate the real achievable spectral-efficiency spurious events or actions. Selective memory can also be gains using actual RF spectrum in a controlled interference used toplaceagreaterbias on recentknowledge rather than and

user-activity

environment.

The cost function determin-longer-term knowledge, or vice-versa. Weighting factors, ing the real increase in spectrum-usage efficiency can be

analogoustofilter coefficients are used toimplementmem- reconfigured

permitting

the exploration of

many

different

ory

selectiveness. The selective nature isreconfigurable by case studies. Examples ofthese include investigating the

varying

the weighting factors associated with the knowl- cost of rapid spectrum allocation where processing power

edge set stored in each

memory

delay. Equation 1 is a is the

determining

factor andinvestigating the cost of

op-conceptual example of this process where y(k) represents

portunistic

access of

narrow

spectrum segments where in-the

kth

desired parameter value/solution,

v(n)

is the

nrh

terference may be the

important

factor.

weightingfactorassigned bythereasoningengine,

x(n)

is A

cognitive

radio faced witha

developing

wireless com-the

nth

knowledgesetelement stored in the memorydelay- m_{munications scenario may be forced}s m b f t e_to

expend

c_considerable line and_{line and}Ndelays

Ndy,isthememory-length.

is the memory-length. The two_{The twoextreme}extreme..._{energy using} _{resources on} _determining_{the best course of}

cases of 1. a memory-less radio device is achievable by action. A better

approach

istobreak down the

developing

assigningaweightingfactors ofzerofor all memorydelay- situation and

apply

asequenceof less

complex

incremental

lineweightsand 2. aphotographicmemory isachievableby solutions. The

objective

inthiscase istosolve the

complex

assigning

a

weighting

factor ofone for all memory

delay-

overall

problem

using

a

combination

of these

incremental

line

weights

and

averaging

the result.

solutions.

The

possibly

complex

scenario

can decomposed

Ndelays into two main

classes,

where incremental solutions for each y(k)

E

w(T)sr(n)

(1)

stage

may

already

exist in the

delay-line

of

knowledge

sets.

n=O The firstcaseisa

repeatable

scenarioand the secondcase

is auniquescenario.

3.2.

Decision

Makling,

Learning and Con-

A repeatable scenario is where similar wireless

com-ciuslons Format'ion

munications tasks andobservable environmental conditions occur more than once. In this case, the cognitive radio,

The mainobjectiveof the decisionmaking, learning and which identifies the emerging similarities from the knowl-conclusions formation element of thereasoningwrapperis edge sets, can invoke asequence ofpreviously successful toproducean

'intelligent'

andtimelyanswerto aproblem procedures in an attempt to accelerate the completion of setbasedonpreviousactions and consequences, currentob- communicationstask(s). Thepotentialbenefits of this abil-servations and

o'bjectives

and

descriptions

of thedata-

ples

ity include

conservationl

of radioresources, increased

(5)

Table 1.

Knowledge

Set

Example

Observations fJ Tasks Actions El Policies

BW: 2MHz Freq: 2.08GHz Voice: High quality OFDM:500 usershare Interference Avoidance

Avail.

eaergy:high

Reconfigure: OFDM TXPower: mm. Social: extrovert

Users: 2 Maintain link Sensebefore use

Mobility: mobile

The unique scenario is where

internal and external ob-

References

servations, available radio resources and desired

communi-cations task are not the same (or in the same sequence) as [1]

Mitola,

J.,

III; Maguire,

G.Q.,

Jr., "Cognitive ra-previously experienced by the cognitive radio. It is possible dio:

making

software radios more personal," Personal that theunique scenario may be decomposed as a sequence

Communications,

IEEE [see also IEEE Wireless

Com-ofrepeatable scenarios. Thus the cognition wrapper can at-

munications],

vol.6,

no.4pp. 13-18, Aug 1999 tempt acombination of proven tactics in order toprovidea

solution.

[2]

Hlaykin

S., Cognitive Radio:

Brain-Empoweredd

The knowledge gained from a communications task may WirelessCommunications, IEEEJournal on Selected include the data-type involved, powerand duration of the AreasinCommunications,February,2005.

transmission, structural, component and parameter config- [3]

http://www.w3.org/XML/

uration of the radio that successfully (or unsuccessfully)

completed the task. The consequences of this communi-

[4]

http://www.w3.org/TR/owl-features/

cations task can include Bit Error Rates (BER),

Quality

of

Service(QoS), consumed radio resources, interference ex-

[5]

J. Neel, J. Reed, R. Gilles. The Role of Game

The-periencedorinflicted, and an estimate of the spectrum effi- ory in theAnalysisof Software Radio Networks, SDR

ciency. ForumTechnicalConference

November,

2002.

An action set is the list of required components, struc-

[6]

Rieser,

CJ;

Rondeau,

T.W.;

Bostian,

C.W;

Gal-turalconfigurations orpossible reconfigurationinstructions l

for existing structures, component parameter-values and

land

Testi

"oa

ribt

teneti

orithm

baise

deadlines by whichthe action set should beimplemented. g gg

Action sets

aepouebycognitive

engine forprogrammable radios," Military

mechansesms

_{mechansms emloyed n the ognitie engie.}

employed

bythe

rening

and

Thehet-IEFF,

leai

Communications_{vol.3, no.pp.}

Conference,

1L437-

₁₄₄₃2004. MILCOM 2004.

Vol.

_3,

31L

_{Oct.-3 Nov.}

eromorphic reconfigurable radio core is instructed to exe- 2004 cute an action set using Control Logic. Control Logic is

themeansofaddressingeach structure, component and pa-

[7]

Hopfield,

J.J., "Artificial

neural networks," Circuits

rametercapable ofbeing implemented by the reconfigurable andDevices

Magazine,

IEEE,vol.4,

no.5pp.3-

10,Sep core. The actionsetforanOFDMreceivercaninclude the 1988

entire setofrequiredcomponents, or incremental changes

to anexisting structure, requiredparametervalues,and the [8] Klir, G.J.,

"Fuzzy

logic," Potentials, IEEE , vol.14,

time-spaninwhichtoimplementthe actionset. no.4pp.

10-1L5,

Oct/Nov

1995

[9] P. Mackenzie, "Reconfigurable Software Radio Sys-4.

Conclusions:tems",

Ph.D

dissertation,

Trini

College Dublin,

Ire-land,2004. Thispaperhaspresentedacognitive wrapper

encapsulat-ing

a

highly reconfigurable

radio core. The influence of this [10] Nolan, K.E.,

Reconfigurable

OFDM

Systems,

Ph.D

wrappercanbe varied in orderto

vary

the

'intelligence'

of thesis,

University

of Dublin,

Trinity College,

May

the

cognitive

radio between thetwo extremes ofabaseline 2005.

radio,

anda

highly-complex

and evolvedsystemto account for scenarios where

complex cognition

may notbe neces-sa

O'bservations,

actions and conclusions developedby

the cognitive radio can be stored in avariable-length

mem-oydelay-lLine.

A selective

mrelmo

mechanis

elnable

the