station Greenhouse microclimatic controlled environment by a mobilemeasuring NJAS - Wageningen Journal of Life Sciences

(1)

ContentslistsavailableatScienceDirect

NJAS - Wageningen Journal of Life Sciences

j ou rn a l h o m e p a g e :w w w . e l s e v i e r . c om / l o c a t e / n j a s

Greenhouse microclimatic environment controlled by a mobile measuring station

Goran Martinovi ´c

^a,∗

, Janos Simon

^b

aFacultyofElectricalEngineering,J.J.StrossmayerUniversityofOsijek

bSuboticaTech,DepartmentofInformatics

a r t i c l e i n f o

Articlehistory:

Received16January2012

Receivedinrevisedform16February2014 Accepted8May2014

Availableonline7June2014

Keywords:

Distantmonitoring FAHP

Greenhouse Mobilerobot

Wirelesssensornetwork

a b s t r a c t

Thispaperinvestigatesagreenhousemicroclimaticenvironmentcontrolledbyamobilemeasuring stationwiththeaimofimprovingperformancebyusingwirelesssensornetworks(WSN)technology.The algorithmsforthemobilemeasuringstationthatperformnavigationtasksarecalledBugalgorithms.The existingpotentialﬁeldmethodbasedalgorithmsareimprovedwithanRSSIsignalpropagationmodel andimplementedonatwo-wheeldrivenrobotdevelopingsystemandtheirperformancesaremeasured andanalyzed.TheimplementationpartisdoneonBoe-BotequippedwithSunSPOTenablingthewireless controlability.ThecontrolsurfaceisgenerallymadewithLabVIEWandarelationaldatabase.Control strategyselectionsystemissupportedbyFuzzyAnalyticHierarchyProcess(FAHP).Navigationofthe mobilemeasuringrobotcanbedonemanually,relyingonthevisualdatafromtherobot’scamera,orit canbeswitchedtoautomaticmodewherethedevelopedalgorithmdoesthenavigationjob.Mobilerobot navigationisbasedonthepotentialﬁeldmethodconsideringawiderangeofenergy-awareparameters.

1. Introduction

Theaimofthispaperisoptimizationofenergyconsumption duringoperationof greenhouses. Theproposedsystemhasthe option of gatheringand monitoring climateparameters related to the microclimatic environment of plants, both inside and outsideof themicroclimatic environmentusingwireless sensor networks.Theareaofresearchwithinthecontrolledmicroclimate environment is an autonomous mobile measuring station that collectsclimatedatawithintheenvironment.Forthemovement oftherobotduringmeasurementsinadynamicenvironment,an improvedmethodofnavigating themobilemonitoringstations wasdevelopedbasedonthepotentialfieldmethod.Mobilerobot navigationisbasedonthepotentialfieldmethodincombination withthereceivedsignal strength ofthe WSN(Wireless Sensor Networks)usedasmarkerstoguidetherobot.Thecombination oflocalizationusingsignalstrengthofWSNnodesandmethodsof potentialfieldsleadstothepossibilityofapplicationofthismethod in anunknowndynamic environment. From thestarting point, themobilerobotshouldfindapath tothetarget ina dynamic

∗Correspondingauthor.Tel.:+38531495401;fax:+38531495402.

E-mailaddresses:goran.martinovic@etfos.hr(G.Martinovi ´c),simon@vts.su.ac.rs (J.Simon).

environment,avoidinganyobstacles.Oneofthecontributionsof researchistoimprovetheperformanceofmobilerobotnavigation in an unknownenvironment. In theinterest of optimalenergy consumptionduringoperationofgreenhouses,anewmodelofan expertsystemformanagingmicroclimatemulti-criteriadecision makingbasedontheapplicationoffuzzyrulesis proposed.Six control strategies were developed for managing greenhouse climateanddependingonthemeasuredclimaticconditions,the choiceofanoptimalcontrolstrategyshallbemadebytheenergy consumptionparameter.Afterashortintroduction,ChapterIIgives anoverviewofotherresearchers’solutionsandexperimentaldata.

ChapterIIIdescribestheenvironmentrealizedforthepurposeof testingthecurrentproject. ChapterIVgivesanoverview ofthe proposedcontrol strategy selection systembased ontheFuzzy AnalyticHierarchyProcess(FAHP).ChapterVshowsanevaluation oftheexpertsystemusedincontrolstrategyselection.ChapterVI presentsexperimentalresultsandanalysisofcollecteddata.Chap- terVIIshowsadiscussionaboutthemobileautomatedgreenhouse controlsystem.WeconcludedChapterVIIIwithanoverviewofthe productioncyclesoftheimplementedgreenhouseenvironment.

2. Relatedwork

Thenavigationmethodcouldmanage therobustmechanism guidingtherobotalongthenear-shortestpathinstaticordynamic http://dx.doi.org/10.1016/j.njas.2014.05.007

(2)

62 G.Martinovi´c,J.Simon/NJAS-WageningenJournalofLifeSciences70–71(2014)61–70 environments,and nofullyinformed map isneeded. Thereare

fewgroupsdoingresearchworkinagreenhousecontrolsystem, e.g.[1],[2]and[3].Theyusevariousdevicesforairtemperature, relativehumidityandsoiltemperaturemeasurementswithawire- lesssensornetwork.Theyhavealsodevelopedaweb-basedplant monitoringapplication.Agreenhousegrowercanreadmeasure- mentsovertheInternet,andanalarmwillbesenttotheowner’s mobile phone by SMS or GPRS if some measurement variable changesrapidly.Inthetestenvironment,sixnodesaredeployed intotworows12.5mapartfromeachother.Onemeshnodeworks asarepeaterandimprovesthethroughputofcommunication.A bridgenodegathersdatafromothersensornodes,whichtransmit temperatureandrelativehumiditymeasurementsinone-minute intervals.Our implementation differsfromother developments intheuseofamobilemeasuringagent,whichprovidesflexibil- ity and robustnessfor the system.Autonomousrobots maybe operatedbydifferentnavigationschemes.Thebestresultswere givenbyamodifiedimplementationofthepotentialfieldmethod.

Robotcontrolarchitectureprovidesrules,guidingprinciplesand constraintsfororganizingarobotcontrolsystem,programsand controlalgorithmssoitcanbeautonomousandachievegoals[4].

TherequirementsfeaturingaWSNareexpectedtosatisfyineffec- tiveagriculturalmonitoringconcernbothsystemlevelissueslike unattendedoperation,maximumnetworklifetime,adaptabilityor evenself-reconﬁgurabilityoffunctionalitiesandprotocolsandthe ﬁnaluserneedse.g.communicationreliabilityandrobustness,user friendly,versatileandpowerfulgraphicaluserinterfaces.Serodio [5]developedandtestedasimilardistributeddataacquisitionand controlsystemformanagingasetofgreenhouses.Severalcommu- nicationtechniqueswereusedfordatacommunications.Atalower supervisionlevel,insideeachgreenhouse,aWLANnetworkwitha radiofrequencyof433.92MHzwasusedtolinkasensornetwork toa localcontroller. Acontroller areanetwork (CAN)waspro- videdtolinkanactuatornetworktothelocalcontroller.Through anotherRFlink(458MHz),severallocalcontrollerswereconnected tothecentralcomputer(PC).High-leveldatacommunicationwas providedthroughEthernettoconnectthecentralPCtoaremote network.Feng[6]implementedawirelessdataacquisitionnetwork tocollectoutdoorandindoorclimatedataforgreenhouses.Sev- eralsolar-powereddataacquisitionstationswereinstalledindoor andoutdoortomeasureandmonitorclimatedata.RFlinkswere establishedamongmultiple(upto32)SPWASsandabasestation, whichwasusedtocontroltheSPWASsandtostorethedata.Liu andYing[7]reportedagreenhousemonitoringandcontrolsys- temusingBluetoothtechnology.Thesystemcollectedenvironment

datafromasensornetworkinagreenhouseandtransmittedthe datatothecentralcontrolsystem.Mizunuma[8]deployedaWLAN inafarmﬁeldandgreenhousetomonitorplantgrowthandimple- mentedremotecontrolfortheproductionsystem.Theybelieved thatthistypeoftheremotecontrolstrategycouldgreatlyimprove productivityandreducelaborrequirements.

3. Workingenvironment,technologyandmethods

TheapplicationsforWSNsaremanyandvaried.Theyareusedin commercialandindustrialapplicationstomonitordatathatwould bedifficultorexpensivetomonitorbyusingwiredsensors.They couldbedeployedinwildernessareas,wheretheywouldremain formanyyears(monitoringsomeenvironmentalvariable) with- outtheneedtorecharge/replacetheirpowersupplies.Theycould formaperimeteraboutthepropertyandmonitortheprogression ofintruders(passinginformationfromonenodetothenext).There aremanyusesofWSNs.TypicalapplicationsofWSNsincludemon- itoring,tracking,andcontrolling.Someofthespecificapplications arehabitatmonitoring,objecttracking,nuclearreactorcontrolling, firedetection,trafficmonitoring,etc.Inatypicalapplication,aWSN isscatteredinaregionwhereitismeanttocollectdatathroughits sensornode.TheWSN-basedcontrollerhasallowedaconsiderable decreaseinthenumberofchangesinthecontrolactionandmade astudyofthecompromisebetweenquantityoftransmissionand controlperformancepossible.Fig.1showsourgreenhousetest- ingenvironment. Thegreenhouseprotectsplantsfromextreme weatherconditions.However,iftheperiodofdaylightprevents thephotosyntheticactivity,theplantsdonotgrow[9].Horticul- turallightingallowsthegrowertoextendthegrowingseason.It enablesayear-roundproductionofplantsormakesitpossiblefor thegrowertostartsowinginearlyspringandcontinueseasonuntil thefirstfrost[10].Plantsneedabout10-12hoursoflighttoimprove growth.

Motioncontrolofmobilerobotsisaveryimportantresearch ﬁeldtoday,becausemobilerobotsareaveryinterestingsubjectin bothscientiﬁcresearchandpracticalapplications.Inthispaper,the objectofremotecontrolistheBoe-Botvehicle.Thevehiclehastwo drivingwheelsandtheangular velocitiesofthetwowheelsare controlledindependently.Whenthevehiclemoves towardsthe targetandthesensorsdetectanobstacle,anavoidingstrategyis necessary,asin[11],[6]and[12].Thehostsystemisconnected tothemobilerobotwiththeSunSPOTmodule.Aremotecontrol programhasbeensupportedbygraphicaluserinterfaceshownin Fig.2.Astothegreenhouseclimatecontrolproblem,thesystemhas

Fig.1.Cropsinthegreenhouse.

(3)

Fig.2.Thedevelopedcontrolcenterwindow.

providedpromisingresults[13].Thisapplicationmonitorsinternal andexternalparametersofthegreenhouse,givesanoverviewof theactuator’scurrentstate,andhasavideosurveillanceoption.

Allthecollecteddatafromthemobilemeasuringstationandfrom theexternalmeteorologystationisstoredindatabase.Navigation ofthemobilemeasuringrobotcanbedonemanually,relyingon thevisualdatafrom therobot’s camera,or canbeswitchedto automaticmodewherethedevelopedalgorithmdoesthenaviga- tionjob.Deploymentofwirelesssensorsandsensornetworksin thegreenhouseenvironmentismostlyatthedevelopmentstage.

Theaforementionedapplicationscanbeclassiﬁedintoﬁvemain categories:

•Environmentalmonitoring,

•Precisionagriculture,

•Machineandprocesscontrol,

•Buildingandfacilityautomation,and

•Traceabilitysystems.

Theimplementedwirelesssensornetworkwasdevelopedby SunMicrosystemsInc. to provideenvironmental monitoringin ﬁelds.Aremoteapplicationservercanrelaydatafromthesen- sornetworktolocalusersviaaWLANandtoremoteusersviaa cellularnetworkandtheInternet.Wehavedevelopedawireless prototypesystemtoacquire,store,displayandtransmitreal-time environmentaldatabetweenthegreenhouseandremotelocation.

Inthecontrolsystem,amobilemeasuringagentnavigatesbetween thecropsandcollectsdata.Nodescommunicatewitheachotheror withagreenhouseserverthroughaWSN.Theservercollectsallthe informationreceivedfromthemotesandstorestheminadatabase.

Testsdemonstratedagreatpotentialtoimproveefficiencyandpre- cisionforgreenhousecontrolanddatacollection.Duringthepast fewyears, potentialfield methods(PFM)forobstacleavoidance havegainedincreasedpopularityamongresearchersinthefield ofrobotsandmobilerobots,asshownin[14],[15]and[16].The ideaofimaginaryforcesactingonarobothasbeensuggestedby [17].Intheseapproaches,obstaclesexertrepulsiveforcesontothe robot,whilethetargetappliesanattractiveforcetotherobot.The sumofallforces,theresultantforce,determinesthesubsequent directionandthespeedoftravel.Accordingto[18]and[19],oneof thereasonsforthepopularityofthismethodisitssimplicityand efficiency.

Themeasuredvaluesofthegreenhouseclimatevariablesare ﬁrstconvertedfromanalogtodigitalonesandthentransmitted

tothecontrolsystem.Becauseofthehighmoistureinthegreen- house,thecontrolsystemisnormallylocatedoutside.Theoptimal greenhouseairtemperaturedependsontheintendedlevelofthe photosyntheticactivity.Eachplantspecieshasitsownoptimalval- uesofairtemperatureandactiveradiationoflight,whichenable themaximumphotosyntheticactivity.Soiltemperaturealsoplays animportantrole.Conductionheattransfersdirectlytothesoil structure and through convection betweentheplant roots and waterﬂowaroundthem[20].Themainconcerninhumidityand temperaturecontrolistoprovidethebestconductivitytoactive movementofwaterandnutrientsthroughtheplant.Humiditycon- trolisalsoanimportanttooltopreventthespreadofplantdiseases ingreenhouses.Fig.3showsadatafusionstructureingreenhouse environmentwithimplementedsensorsandactuatorsresponsible forclimateconditioning.

The range sensors assist only for wall following and wall detecting purposes. For a greenhouse environment, the mobile measuringagentisequippedwithAgriculturalSensorandGasSen- sorboards,andtheycanbeusedtogethertomeasureandcontrol airhumidity,CO₂levelandairtemperature.UsingaPAR(Photo- syntheticallyActiveRadiation),thesensorcheckstheconditions forphotosynthesis[21].Fig.4showsthecompletecontrolsystem ofthegreenhouse,similarlyto[22].Theactuatorpart,likeheaters, foggers,coolersofthesystem,iscontrolledbyaLabVIEWapplica- tion.Allnecessaryactuatorsareusedtocreatearelativelyconstant environmentalconditioninsidethegreenhouse.Ifthetemperature rises,thecontrolsystemenablescoolingfansandfoggersuntilwe getanacceptabletemperature.Ifthetemperaturefallsbelownor- mal,oilheatersactivate.Themobilemeasuringagentdoesnotneed tomonitorvaluescontinuouslyanditcanspendlongperiodsin apowersavingmode.Inourcase,themeasuringtouracrossthe greenhouseismadeonceperhour.Thecollecteddataistransmitted toarelationaldatabaseforfurtherstatisticalanalysis.Usingsensor dataautomaticallyadjuststhecontrolsystemtomatchlocalcondi- tionsinthegreenhouse.Avoidingover-wateringalsohelpsprevent certaincropdiseasesincludingrot,fungiandbacteria,whichthrive inwetconditions.

4. FuzzyAHPapproachtochooseacontrolstrategy

Aswithmostissues related todecisionmaking, selectionof thecontrolstrategybecomescomplicatedinarealenvironment.

Thereisahighprobabilitythatthehumanfactoriswronginmak- ingdecisions orpredictionsfor quantitativeproblems.In many

(4)

64 G.Martinovi´c,J.Simon/NJAS-WageningenJournalofLifeSciences70–71(2014)61–70

Fig.3.Datafusionstructureinthegreenhouseenvironment.

cases,it tends toexpress thecurrent state of thesystemwith verbalexpressions.Afuzzylinguisticmodelallowsmappingofa verbalexpressiontoanumericvalue.A multi-criteriadecision- makingmethodbasedonfuzzyrelationsisusedforquantitative determination of the importance of each criterion with some degreeofinaccuracy.Inthiscase,theFAHP(FuzzyAnalyticHier- archyProcess)multi-criteriaanalysisisproposedasatoolforthe implementationofamulti-criteriascheme.AHPisamethodfor decision-making,whichissuitableforsolvingcomplexunstruc- turedandmulti-attributeproblemsdevelopedbyT.L.Saaty[23].

Themostcreativepartinmakingdecisionsthatgreatlyaffectthe outcomeofthedecisionisproblemmodeling.Identificationofthe hierarchyofdecisionsisakeyfactorintheapplicationoftheAHP method.AHPisthebasisfortheformalizationofcomplexproblems usingthehierarchicalstructureoftheapplicationbycomparing pairsofattributes.Thismethodiswidelyappliedinindustrialappli- cationsandotherareas.ConventionalAHPisnotabletoreflectthe humanthinkingstyle.Forthisreason,FAHPisdevelopedtosolve thehierarchicalfuzzyproblems.IntheFAHPmethod,allcalcula- tionsareperformedwithfuzzynumbers,asproposedin[24].In thispart,theFAHPmethodisconsideredfortheselectionofcon- trolstrategiesinmanagingmicroclimateconditionsapplyingthe multi-criteriadecisionmakingprocess.Forthepurposeofdefining afuzzyset,letXbethedefinedarea, ÃafuzzysubsetofXsuchthat itappliestoallx∈X.LettherebeanumberA˜(x∈[0,1])assigned torepresentthemembershipofxto ÃandletA˜(x)becalledthe membershipfunctionof Ã.Afuzzynumber Ãisanormalandconvex fuzzysubsetofX.Accordingto[25],aconvexsetimplies(1):

∀x1∈X,x2∈X, ∀˛∈[0,1]

A˜(ax1+(1−a)x2≥minA˜(x1),A˜(x2))

(1)

Atriangularfuzzynumber ˜Acanbedeﬁnedbyatriplet(a,b,c).

Themembershipfunctioncanbedeﬁnedasin[25]:

A˜(x)=

⎧ ⎪

⎪ ⎪

⎨

⎪ ⎪

⎪ ⎩

x−a

b−a, a≤x≤b, c−x

c−b, b≤x≤c,

0 ·

(2)

Wellknownaddition,multiplication,subtractionanddivision oftriangularfuzzynumbersaredescribedin[25]andtwoofthese operationsusedinthispapercanbeexpressedasfollows:

Additionoffuzzynumbers⊕:

(a1,b1,c1)⊕(a2,b2,c2)=(a1+a2,b1+b2,c1+c2) (3) Multiplicationoffuzzynumbers⊗:

(a1,b1,c1)⊗(a2,b2,c2)=(a1×a2,b1×b2,c1×c2) (4) TheFAHPmethod isa systematic approach tothechoiceof alternativesandjustiﬁcation oftheproblemusingtheconcepts offuzzysetsandanalysisofhierarchicalstructures.Thedecision makercanspecifythesettingsintheformofthenaturallanguage oranumericvalueontheimportanceofeachattribute.Thesys- temcombinesthesesettingswithexistingdatausingtheFAHP method.IntheFAHPmethod,pairwisecomparisonsinthematrix arefuzzynumbersandfuzzyarithmeticoperators.Theprocedure calculatesthesequenceofweightvectorsthatwillbeusedforthe selectionofthemainattributes.Insomecases,thedecisionmaker candeterminetheimportanceinastandardformwithpairwise comparisonsdeﬁnedbyT.L.Saaty[23]intheformofninepoints onthescaleofimportancebetweenthetwoelements.Triangular fuzzynumberswereintroducedintheconventionalAHPinorderto improvethelevelofjudgmentofdecisionmakers.Thecentralvalue ofthefuzzynumberisthecorrespondingactualvalue.Expanding thenumber isanestimateoftheactualnumber.Ifthedecision makerisunabletospecifytheirpreferencesaccordingtonumeri- calvalues,itisalsopossibletospecifythesettingintheformof natural languageexpressionsof theimportanceof each perfor- manceattributes.Thedecisionmakeralsousesthefuzzymethod fortheconstructionofthelookuptableandtheappropriatevalue offuzzynumberssimilarasin[24].In theFAHPmethod,using fuzzyarithmeticandaggregationoperators,theprocedurecalcu- latesthesequenceofweightvectorsthatareusedtodetermine theimportanceofeachattribute.TherearemanyvariationsFAHP methodsandimprovementsproposedbyseveralresearchers.In thenextfewsteps,themethodanalysiswillbegivenandthenthe methodwillbeappliedtotheproblemofselectingtheoptimal controlstrategy.

Step1:Intheﬁrststep,theFAHPmethoddevelopsthehierarchi- calstructureoftheproblemsimilarto[23].Afterthat,thedecision

(5)

Fig.4.Ahierarchyfortheselectionofthecontrolstrategy.

makermustdeterminerelative weightingfactorsforeachcrite- rion.WiththeAHPmethod,weightingfactorsaredeterminedby pairwisecomparisonofeachcriterion.Todeterminetherelative weight,thedecisionmakerisaskedtomakepair-wisecompari- sonbyusingascalefrom ˜1−9.˜ Datafrompairwisecomparisonis organizedintheformoftriangularfuzzynumbers.

Step2:Ifthedecisionmakercannotusepreferencesformsof triangularfuzzynumbers,thereisapossibilityofusinglinguistic termsbyapplyingthelookuptablefromwhichthecorresponding valuescanbereadilyextractedforfuzzynumbers.

Step3:Aftersettingupthehierarchyandpairwisecomparison ofthecriteriaandalternatives,itisnecessarytocalculatetheglobal valueofpriorityofalternativesaspresentedin[23].

Inordertodeterminetheoptimalstrategyformanagingmicro- climateconditions,ahierarchicalmodelisdevised,asshownin Fig.4.Inthiscase,theFAHPmethodisusedfordeterminingthe optimalcontrolstrategybasedonthemeasureddata.

Theﬁrstlevelisthegoalitself.Inthiscase,thegoalistodeter- mine theoptimalcontrol strategy. Thegoal isdivided intothe

following three main criteria: A - mobile measuringstation, B -outdoorweather station,and C -indoorstationarymeasuring point.Thethirdlevelincludessystemparameters.Improvedalgo- rithmicstepsoftheproposedapproach from[25] areshownin Fig.5.

Categoryone–Themobilemeasuringstation:Generalterms andconditionsinsidethemicroclimateenvironmentaremeasured by usinga mobilemeasuringstation equippedwiththeneces- sarysensors.Datacollectedfromtheenvironmentareforwarded tothecentraldatabasewiththeexactlocationandtimeofmea- surement.Inﬂuencingfactorsaredividedintosixsub-criteria:A1 -CO2gasconcentration,A2–Airhumidity,A3-Airtemperature, A4-Atmosphericpressure,A5-Solarradiation,A6-Positionofthe measurement.

Categorytwo–Outdoorweatherstation:Tocontrolmicrocli- mateconditions,itisnecessarytomonitorexternalclimatefactors aswell.Datacollectedfromtheoutsideweatherstationsarealso placedinthedatabase.Takingintoaccountexternalfactors,the secondcriterionisdividedintosixsub-criteria:B1-Humidity,B2

(6)

Fig.5.ImprovedalgorithmicstepsoftheFAHPapproach.

-Airtemperature,B3-Theamountofrainfall,B4-Solarradiation, B5-WindSpeed,B6-Winddirection.

Categorythree–Indoorstationarymeasuringpoint:Theindoor stationarymeasuringpointisintroducedforthepurposeofimprov- ingcontrolperformance.Besidesstandardinternalfactorssuchas temperature,relativehumidity,lightingconditions,etc.,wemea- surethesoiltemperatureandsoilmoisture.Thethirdcriterionis dividedintoﬁvesub-criteria:C1-Soilmoisture,C2-Soiltemper- ature,C3-Airtemperature,C4-Airhumidity,C5-Solarradiation.

Finally, the fourth and ﬁnal level contains alternatives. Six controlstrategiesweredeveloped:STR1–DayHighPerformance, STR2 – DayNormal, STR3 – DayEconomic, STR4 – NightHighPerformance, STR5 – NightNormal, STR6 – NightEconomic.

Theselectionoftheoptimalcontrolstrategyisbaseduponinput factors.ProceduresofFAHPcalculationsaregiveninthefollowing way,similarasin[23]:

Procedure1–Inordertodeterminetheoptimalcontrolstrat- egy,ahierarchicalstructureiscreatedasshowninFig.4.Taking intoconsiderationtherequirementsofthegoal,thedecisionmaker playsakeyroleintheevaluation,evaluatingtheresultsofpair- wise comparisonoftheﬁrstlevel ofthehierarchy.Byapplying triangularfuzzynumbers,usingpairwisecomparison, thefuzzy judgmentmatrixisconstructedas ˜A(i,j),wherea^−˛_ij =˜1,3˜,5˜,7˜,9˜ or ˜1⁻³,3˜⁻¹,5˜⁻¹,7˜⁻¹,9˜⁻¹ifiisnotequaltoj,similarto[23].Mem- bershipfunctionsareperformedbyusingan˛cut.The˛cutplays theroleofunifyingreliabilitypropertiesofexpertsanddecision makersduringthejudgmentprocess.Thiswillgiveasetofvalues

(7)

Fig.6.Constructionofthefuzzyjudgmentmatrix.

intheintervaloffuzzynumbers.Thelowerlimitandupperlimit offuzzynumberswithrespecttothe˛cutaredefinedbyequation describedin[23]. ãij=[aij,bij,cij]isoneoftheelementsof Ãwith aclosedintervalwhosemidvalueisbij.Thenbijisjustoneofthe integersfromonetonine,whichareusedinthemethodofAHPin [26]and[27].

Accordingto[25],whenıisselectedtobelessthan1/2,bijis selectedastheconsecutivetwo-levelscalemidpoint,anddisthe crossoverpointoftwotriangles(Fig.6).If(d)iszero,thereisno impactontheentiredistinctcognitive-fuzzyconversion.Ifıhas avaluegreaterthanone,theleveloffuzzinessincreases,butthe degreeofconﬁdencedecreases.Itisproposedtoselectavaluefor ıbetween1/2and1,asshowninFig.6.

Afterpairwisecomparisonofallelements,matrix ˜Aisconverted intofuzzytriangularnumbers,andthegeometricmeanmethodis appliedtocalculatetheprioritiesofthesetriangularfuzzynumbers [23],asin(5)and(6).

k˜_i=

a˜_ij

1/n

i=1,...,n (5)

Valueof ãijcanbedefinedas ãir1j,a˜ir2j,a˜ir3j:

kir1= air1,j

1/n

k_ir2= a_ir2,j

1/n

k_ir3= a_ir3,j

1/n

⎫ ⎪

⎪ ⎪

⎬

⎪ ⎪

⎪ ⎭

j=1,...,n (6)

Accordingto[23],foreachofthealternativesorcriteria,weights canbecomputedasfollowsin(7):

q_ir1=

k_nir1 i=1k_ir3

q_ir2=

k_n_ir2 i=1k_ir2

q_ir3=

k_n_ir3 i=1k_ir1

⎫ ⎪

⎪ ⎪

⎬

⎪ ⎪

⎪ ⎭

i=1,...,n (7)

Fig.7showsusualandknownconversionprocedureofalinguis- ticvariableintofuzzynumbers,asdescribedin[25].

After that, the weight of criteria i can be written as ˜qi= (qi,r1,qi,r2,qi,r3).Thegivenweightsareintheformoftriangular fuzzynumbers.Thedefuzziﬁcationprocess[13]isdoneaccording to(8):

qi= q˜_i

3 (8)

Fig.7. Membershipfunctionsoflinguisticvaluesforcriteriarating.

Table1

Deﬁnitionoffuzzymembers.

Fuzzylanguage Fuzzyvalues Meaning

VB (1,1,2) VeryBad

VP (1,2,3) VeryPoor

P (2,3,4) Poor

F (3,4,5) Fair

G (4,5,6) Good

VG (5,6,7) VeryGood

I (6,7,7) Ideal

Table2

Valuesoftheﬁrsthierarchylevel.

A B C

A 1 2 5

B 1/2 1 3

C 1/5 1/3 1

Table3

TheevaluationmatrixoftheAcriteria.

A1 A2 A3 A4 A5 A6

A1 1 1/3 2 1/3 1/5 1

A2 3 1 2 2 1/2 5

A3 1/2 1/2 1 1/2 1/3 1

A4 3 1/2 2 1 1/2 3

A5 5 2 3 2 1 5

A6 1 1/5 1 1/3 1/5 1

Table4

TheevaluationmatrixoftheBcriteria.

B1 B2 B3 B4 B5 B6

B1 1 3 1 1/4 1/3 2

B2 1/3 1 1 1/3 1/2 1

B3 1 1 1 1/2 1 2

B4 4 3 2 1 3 3

B5 3 2 1 1/3 1 1

B6 1/2 1 1/2 1/3 1 1

Procedure2–Ifthedecisionmakerisunabletodeterminethe importanceorthepriorityofcriteria,itispossibletouselinguis- ticvariables toestimatetheimportanceofcriteriawithrespect tothegoal,andlinguisticvariablestoestimatetheimportanceof alternativeswithrespecttoeachcriterion[13].Asdescribedin[25], thelinguisticvariablecanbeeasilyconvertedintofuzzynumbers accordingtostandardprocedurebyusingFig.7andTable1.

Procedure3–Thedegreeofimportanceofeachobjectivecanbe incorporatedintotheformulationbyapplyingfuzzyprioritiesand evaluatingalternatives.Aweightedpriorityofeachalternativecan beobtainedbymultiplyingtheevaluationmatrixbyvectorweights andbysummingofallattributes.Tables2–5showtheevaluation matrixrelevanttoindividualcriteriaingreenhouseenvironment.

Todeterminetherelativeweight,thedecisionmakerisaskedto makepairwisecomparisonbyusingascalefrom ˜1−˜9similarto

Table5

TheevaluationmatrixoftheCcriteria.

C1 C2 C3 C4 C5

C1 1 5 7 2 3

C2 1/5 1 1 1/3 1/3

C3 1/7 1 1 1/5 1/5

C4 1/2 3 5 1 1

C5 1/3 3 5 1 1

(8)

Table6

Linguisticevaluationsofalternativesaccordingtocriteria.

Alternatives Criteria

A1/B1/C1 A2/B2/C2 A3/B3/C3 A4/B4/C4 A5/B5/C5 A6/B6

STR1 G/G/F F/G/G VG/F/G G/F/VG G/VG/VG G/VP

STR2 VG/F/G F/G/G F/G/I VP/G/G F/G/G VG/F

STR3 VP/VG/F VG/F/F VG/VP/G G/G/F VG/VG/F VG/G

STR4 F/G/G F/G/VG VP/VG/F VG/VG/VP G/VG/F G/F

STR5 VG/G/VG I/F/G G/F/G G/F/VG G/G/F VP/F

STR6 VP/G/VG F/F/F VG/G/G I/VP/VP VG/VG/G VG/G

Table7

Comparisonoftwoweightedmethodswithdifferent␦values.

␦=0.6 ␦=0.7 ␦=0.8 ␦=0.9

FAHP Yager FAHP Yager FAHP Yager FAHP Yager

Rank Score Rank Score Rank Score Rank Score Rank Score Rank Score Rank Score Rank Score

STR3 6.24 STR3 0.99 STR3 6.62 STR3 0.98 STR3 7.07 STR3 0.98 STR3 7.61 STR3 0.98

Fig.8.Measuredtemperature.

[23].Datafrompairwisecomparisonisorganizedintheformof triangularfuzzynumbers.

f(aj,qi)=

q˜j⊗a˜j (9)

Oneofthesixalternativecontrol strategiesis chosenasthe optimalstrategyforthegivenconditions.Table6showslinguis- ticevaluationslinkedtospeciﬁccriteria.Afterobtainingtriangular fuzzynumbers,theirpriority(geometricmeanmethod)iscalcu- latedbyusing(5)and(6).

Foreachcriterionoralternative,theweightingfactoriscalcu- latedbyusing(6).Afterdefuzziﬁcationoffuzzyweights,thenew valueofweightingfactorscanbeobtainedbyusing(8).

5. Evaluationoftheexpertsystem

Inordertoshowtheapplicabilityoftheproposedexpertsys- tem,theproposedmethodiscomparedwithYager’smethodunder various␦cuts.Thevaluesof␦cutare0.6,0.7,0.8and0.9forthe evaluationofsixalternativesinFAHPandYager’smethodandthe resultsareshowninTable7.

BothmethodsareimplementedintheMATLABprogramming environmenttoprovideexperimentalvalues.Comparingmodels withbothmethodsundervarious␦cutlevelscanbeconsideredas

adecisionsupportmodelsinceitguidesdecisionmakerstoselect thebestalternative.InthecaseoftheanalysisinTable7,wecan establishthatundergivenconditionsSTR3DayEconomicisoffered asthebestchoiceforall␦sectionsinFAHPandYager’sapproach.

Bothmethodsareinsensitivetothevaluesof␦becausetheeval- uationresultsdonotchangeinthecaseofvarious␦sections.This factprovesreliabilityandconsistencyofbothmethods.Inthecase ofrankingallavailablealternatives,therearecertaindifferences intwomethods.TheadvantageoftheFAHPapproachisthatitis morenaturalandtheinputparametersfortheapplicationofthe fuzzylinguisticvariablearedeﬁnedmoreprecisely,andtherefore thedecisionofanexpertsystemitselfismorereliable.

6. Experimentalresultsandanalysis

Experimentalmeasurementsaredoneinsidethegreenhouse.

Thebasicsurfaceofthemicroclimateenvironmentisalmost16m², the model is built for experimental purposes in the backyard ofSubotica-Tech.25containersarelocatedinsidethecontrolled microclimaticenvironment,andineachofthemthereisoneplant.

Thepossibilityofimplementingafuzzycontrollerusedasasubsti- tutefortheconventionalcontrolsystemintermsofmaintainingthe desiredtemperatureinthemicroclimaticenvironmentthatis21^◦C

(9)

Fig.9.Simulatedmap.

entailsprecisecontrol.Fig.8showsthegraphofmeasuredexternal temperature,internaltemperatureandfromthetemperaturesen- soronthemobilemeasuringagentinaperiod0-24h.Weareable tomanageinternalconditionsinsidethegreenhouseatarelatively constantvalue,whichisrealizedbyusinggoodandrobustcontrol techniques.

Thisresearchconsiderstheinformation-awareoperationofthe nodesandofthenetworkasawhole.Thisinvolvesconsideration ofthe‘information’containedineach packetbyquantifyingthe usefulnessofdatatothecontrolsystem[28].Themobilerobothas oneWSNnodeonboardanditcommunicateswithothernodesin theWSNnetwork.Thismeansthatwehave aWSNgridwitha knownnodepositionandamobilenodeontherobot.Themobile nodeislocalizablebyusingtheRSSIparameterofthethreeclosest nodes.Forthelocalizationofmobilemeasurementstationsamin- imumofthreeWSNnodesisrequiredthatareintherangeofthe measuringstation.Sincethepositionsofﬁxednodesareknown, simplybymeasuringtheRSSIsignalfromthethreenearestnode, wecandeterminetheexactpositionofthemeasuringstation.In [11],detailednavigationalgorithmformobilemeasuringstationis presented.Fig.9showsasimulationmapforthepathlengthcom- parisontest.Fig.10depictsanevaluationofthepathlengthfor everyimplementednavigationalgorithm.

Everyﬁxednodehasatemperatureandhumiditysensor,while themobilerobotisequippedwithtemperature,humidity,light, solarradiation,CO₂concentrationandpressuresensorssimilarto [29].Fixedsensornodesareplacedonevery100metersandthe mobilemeasuringstationimprovesmeasurementswiththeability tomeasureenvironmentalparametersbetweentheﬁxednodes.By

Table8

Theaveragetotalweightandthenumberoffruitharvested.

Testedplants Average weightof fruit(with automation)

Average weightof fruit(without automation)

Average numberof fruitper plant(with automation)

Average numberof fruitper plant (without automation)

Tomato 215g 185g 18 11

Capsicum 140g 120g 17 12

Cucumber 80g 60g 15 12

usingthisconceptwecanbuildatemperature,humidityoreven lightmapoftheobjectinagoodresolutionandwecandetermine ifanymalfunctionappears(likeplasticﬁlmdamageorbio-light malfunction,etc.).

AswecanseeinFig.10,thedevelopedRSSIPotFieldnavigation algorithmﬁnishesthenavigationtaskwiththeshortestpathlength.

7. Discussion

Usinganautomatedgreenhousecontrolsystemitispossibleto expandtheproductioncycleofthecropsinsidethegreenhouse, butwemusttakeintoconsiderationtheenergyusagefactor.Ifwe spendtoomuchenergytocreateidealconditionsforthecrops, theproductioncostswilldrasticallyrise.Thisisthereasonwhyit isnotsuggestedtogrowplantsintheperiodfromDecember15to February15.Excellentlight,moderateheatingcostsandgoodprices annuallydemonstratethisisthebesttimeforgreenhousetomato production.Tomatoplantsgrowbestwhenthenighttemperature ismaintainedat16-18^◦C.Temperaturesbelow16^◦Cwillprevent normalpollinationandfruitdevelopment.Forexample,tomatoes areawarmseasonvegetablecropandtheygrowbestundercon- ditionsofhighlightandwarmtemperatures.Lowlightinafallor wintergreenhouse,whenthereislessthan15%ofsummerlight levels,greatlyreducesfruityieldwhenheatingcostsarehighest.

Forthisreason,itisdifﬁculttorecommendthatagreenhouseoper- atorshouldgrowandharvestfruitfromDecember15toFebruary 15.Basedonafewyearsofexperience,tomatoproductionismost successfulinthespring.

Inwarmorhotoutdoorconditions,tomatogreenhousesmust beventilatedtokeeptemperaturesbelow35^◦C.Hightemperatures notonlyaffecttheleavesandfruit,butincreasedsoiltempera- turesalsoreducerootgrowth.Table8givesacomparisonofthe

Fig.10. Pathlength.

(10)

70 G.Martinovi´c,J.Simon/NJAS-WageningenJournalofLifeSciences70–71(2014)61–70 averagetotalweightandthenumberoffruitharvestedwiththe

developedcontrolsystembasedontheWSNandmobilemeasur- ingstationandwithaclassicalone-pointmeasuringcontrolsystem.

Successingreenhouseplantsdependscompletelyonfruityield.

Yieldsof20–25%gainperplantareverygoodforannualcosts.One ofthemaingoalsoftheprojectistodevelopastabilizeduniver- salgreenhousecontrolsystemcapableoffastadaptivecontrolin variousconditions.Theresearchareainsidethegreenhouseisthe navigationofawirelesslycontrolledmobilemeasuringstation.

8. Conclusion

Inthiswork,weanalyzetheneedforthedevelopmentofmodels forenvironmentalmanagementduetothefactthatthestationary measuringpointscanbereplacedbyanautonomousmobilemea- suringstation.Theseassumptionsimplytheneedforaddressing autonomous navigationof the mobile measuring stationin an unknowndynamicenvironment,aswellaswirelesslycollectedcli- maticparametersbyusingsensornetworks.Inaddition,thereisa needfordevelopinganexpertsystembasedonfuzzymultiplecrite- riadecisionmakingandanoptimalmanagementstrategyofchoice forthegreenhouseaswellasthedevelopmentofmethodsfordata fusioninrealtime.Theproposedmodelforenvironmentalmanage- mentusingmobilemonitoringstationscomparedtoconventional systemshastheabilitytocontrolmicroclimateconditionsbyzones, thepossibility of generating a temperature map of the object, theobjectdamagedetectionbasedonthetemperaturemap.This changeinmanagementmicroclimatesledtocertainchanges,which arereﬂectedinanincreasingnumberofcontrolstrategies,which isoneofthemostimportantelementsinthedevelopedmodel.

Resultsoftheanalysisofthebehavioralcontrolsystemindicated theneedforspecialattentiontoaddresstheverystageofdecision- makingthat bringsa stable microclimate withminimal energy expended.Basedondefinedsubjects,researchincludesthebasic hypothesis,whichisreflectedinthepossibilitiesofusingamobile measuringstationinsteadofstationarymeasurementpoints.The controlsystemhasbeensuccessfullyimplemented;however,there arestillpossibilitiesforfurtherimprovementsmentionedinChap- terVI.Inordertocarryoutverificationoftheproposedmodelfor environmentalmanagement byanautonomousmobilemeasur- ingstation,comparativeresultsofexperimentalmeasurementsof variousclimateparametersaregiven.Theanalysisoftheresults indicatesaccuracyandrobustnessofthemodel.

References

[1]A.Pawlowski,J.L.Guzman,F.Rodríguez,M.Berenguel,J.Sánchez,S.Dormido, SimulationofGreenhouseClimateMonitoringandControlwithWirelessSen- sorNetworkandEvent-BasedControl,Sensors9(1)(2009)232–252.

[2]L.Bencini, D.DiPalma, G.Collodi,A.Manes, G.Manes,“WirelessSensor Networks forOn-Field AgriculturalManagement Process,WirelessSensor Networks: Application - Centric Design”, Yen Kheng Tan (Ed.), InTech, 2010, Available from: http://www.intechopen.com/books/wireless-sensornetworks-application-centric-design/wireless-sensor-networks-for-on-ﬁeld- agricultural-management-process

[3]J.Simon,OptimalMicroclimaticControlStrategyUsingWirelessSensorNet- workandMobileRobot,ActaAgriculturaeSerbica18(36)(2013)3–12.

[4]R.Langer,L.Coelho,G.Oliveira,K-Bug,aNewBugApproachforMobileRobot’s PathPlanning,in:Proc.oftheIEEEInt.ConferenceonControlApplications, Singapore,1-3October,2007,pp.403–408.

[5]C.Serodio,J.B.Cunha,R.Morais,C.A.Couto,J.L.Monteiro,ANetworkedPlatform forAgriculturalManagementSystems,ComputersandElectronicsinAgricul- ture31(1)(2001)75–90.

[6]X. Feng,T. Yu-Chu,S.Yanjun,S.Youxian, WirelessSensor/ActuatorNet- work Design for Mobile Control Applications, Sensors 7 (10) (2007) 2157–2173.

[7]G.Liu,Y.Ying,ApplicationofBluetoothTechnologyinGreenhouseEnviron- ment,MonitorandControl,J.ZhejiangUniv.,Agric.LifeSci.29(1)(2003) 329–334.

[8]M.Mizunuma,T.Katoh,S.Hata,ApplyingITtoFarmﬁelds,AWirelessLAN.NTT Tech.Rev.1(2003)56–60.

[9]R.Siegwart,R.Illah,IntroductiontoAutonomousMobileRobots,TheMITPress Cambridge,London,England,2004.

[10]S.J.E.Mohd,A.H.Adom,A.Y.Shakaff,M.A.Shuib,Real-TimeWirelessAgri- culturalEcosystemMonitoringforCucumusMelo.lCultivationonNatural VentilatedGreenhouse,InternationalJournalofScientiﬁcandResearchPub- lications3(11)(2013)1–6.

[11]J.Simon,G.Martinovi ´c,NavigationofMobileRobotsUsingWSN’sRSSIParame- terandPotentialFieldMethod,ActaPolytechnicaHungarica,JournalofApplied Sciences10(4)(2013)107–118.

[12]C.H.Chiang,J.S.Liu,Y.S.Chou,ComparingPathLengthbyBoundaryFollowing Fast MatchingMethodandBugAlgorithmsforPathPlanning,Opportuni- ties andChallenges forNext-GenerationArtiﬁcial Intelligence214 (2009) 303–309.

[13]L. Gonda,C. Cugnasca,AProposal ofGreenhouseControlUsingWireless SensorNetworks,in:Proc.of4^thWorldCongressConferenceonComput- ersinAgricultureandNaturalResources,Orlando,FL,USA,24-26July,2006, pp.229–233.

[14]O.Khatib,ThePotentialFieldApproachandOperationalSpaceFormulation inRobotControl,in:Proc.of4^thYaleWorkshoponApplicationsofAdaptive SystemsTheory,YaleUniversity,NewHaven,CT,USA,1985,pp.208–214.

[15]M.J.Matari ´c,TheRoboticsPrimer,TheMITPress,London,England,2007.

[16]K.Kreichbaum,ToolsandAlgorithmsforMobileRobotNavigationwithUncer- tainLocalization,CaliforniaInstituteofTechnology,USA,2006,PhDThesis.

[17]O.Khatib,Real-TimeObstacleAvoidanceforManipulatorsandMobileRobots, Int.J.ofRoboticResearch5(1)(1986)90–98(2006).

[18]J.S.Esteves,A.Carvalho,C.Couto,GeneralizedGeometricTriangulationAlgo- rithmforMobileRobotAbsoluteSelf-Localization,in:Proc.of2003IEEEInt.

SymposiumonIndustrialElectronics,RiodeJaneiro,Brazil,09-11June,2003, pp.346–351.

[19]Gy.Mester,WirelessSensor-basedControlofMobileRobotMotion,in:Proc.of 7^thIEEEInt.SymposiumonIntelligentSystemsandInformatics2009,Subotica, Serbia,24-26September,2009,pp.81–84.

[20]J. Simon, G. Martinovi ´c, Web BasedDistant Monitoringand Control for GreenhouseSystemsUsingtheSunSPOTModules,in:Proc.of7thIEEEInt.Sym- posiumonIntelligentSystemsandInformatics2009,Subotica,Serbia,24-26 September,2009,pp.1–5.

[21]Y.Takahashi,T.Komeda,H.Koyama,DevelopmentofAssistiveMobileRobot System:Amos,AdvancedRobotics18(5)(2004)473–496.

[22]I.Matijevics,J.Simon,“ImprovingGreenhouse’sAutomationandDataAcqui- sitionwithMobileRobotControlledSystemviaWirelessSensorNetwork, WirelessSensorNetworks:Application-CentricDesign”,YenKhengTan (Ed.), InTech, 2010. Available from: http://www.intechopen.com/books/

wireless-sensor-networks-application-centric-design/improvinggreenhouse-s-automation-and-data-acquisition-with-mobile-robot- controlled-system-via-wirel

[23]T.L.Saaty,TheAnalyticHierarchyProcess,McGraw-Hill,NewYork,USA,1980.

[24]A.Sriraman,R.V.Mayorga,AFuzzyInferenceSystemApproachforGreenhouse ClimateControl,EnvironmentalInformaticsArchives2(1)(2004)699–710.

[25]T.C.Wang,Y.H.Chen,Applyingfuzzylinguisticpreferencerelationstothe improvementofconsistencyoffuzzyAHP,InformationSciences178(19) (2008)3755–3765.

[26]E.Albayrak,Y.C.Erensal,UsingAnalyticHierarchyProcess(AHP)toImprove HumanPerformance:anApplicationofMultipleCriteriaDecisionMakingProb- lem,JournalofIntelligentManufacturing15(4)(2004)491–503.

[27]O.S.Vaidya,S.Kumar,AnalyticHierarchyProcess:anOverviewofApplications, EuropeanJournalofOperationResearch169(1)(2006)1–29.

[28]V.Lumelsky,A.Stepanov,PathPlanningStrategiesforaPointMobileAutoma- tonMovingAmidstUnknownObstaclesofArbitraryShape,Algorithmica2(1) (1987)403–430.

[29]J.Simon,G.Martinovi ´c,DistantMonitoringandControlforMobileRobots UsingWirelessSensorNetwork,in:Proc.of10thInt.SymposiumofHun- garianResearchersonComputationalIntelligenceandInformatics,Budapest, Hungary,12-14Nov,2009,pp.1–9.