Research Article Performance Analysis of ZigBee Wireless …downloads.hindawi.com › journals ›...

Research ArticlePerformance Analysis of ZigBee Wireless Networks for AALthrough Hybrid Ray Launching and Collaborative Filtering

Peio Lopez-Iturri1 Fran Casino2 Erik Aguirre1 Leyre Azpilicueta34

Francisco Falcone13 and Agusti Solanas2

1Electrical and Electronic Engineering Department Public University of Navarre 31006 Pamplona Spain2Department of Computer Engineering and Mathematics Rovira i Virgili University 43007 Catalonia Spain3Institute for Smart Cities Public University of Navarre 31006 Pamplona Spain4School of Engineering and Sciences Tecnologico de Monterrey 64849 Monterrey NL Mexico

Correspondence should be addressed to Agusti Solanas agustisolanasurvcat

Received 2 September 2015 Accepted 26 January 2016

Academic Editor Pietro Siciliano

Copyright copy 2016 Peio Lopez-Iturri et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

This paper presents a novel hybrid simulation method based on the combination of an in-house developed 3D ray launchingalgorithm and a collaborative filtering (CF) technique which will be used to analyze the performance of ZigBee-based wirelesssensor networks (WSNs) to enable ambient assisted living (AAL) The combination of Low Definition results obtained by meansof a deterministic ray launching method and the application of a CF technique leads to a drastic reduction of the time andcomputational cost required to obtain accurate simulation results The paper also reports that this kind of AAL indoor complexscenario with multiple wireless devices needs a thorough and personalized radioplanning analysis as radiopropagation has a strongdependence on the network topology and the specific morphology of the scenario The wireless channel analysis performed byour hybrid method provides valuable insight into network design phases of complex wireless systems typical in AAL-orientedenvironments Thus it results in optimizing network deployment reducing overall interference levels and increasing the overallsystem performance in terms of cost reduction transmission rates and energy efficiency

1 Introduction

The wide adoption of diverse technological elements inparticular those related to Information and CommunicationTechnologies is a key driver in the transformation provisionand delivery of healthcare services Traditionally healthcareservices entailed large amounts of resources many ofwhich required patients to be in direct contact with healthspecialists for diagnostics as well as for treatment In thepast decade with the steady adoption of software solutionsand seamless connectivity electronic health (e-Health) andmobile health (m-Health) have enabled ambient assistedliving (AAL) and context-aware scenarios In this contextreal time monitoring and interaction of patients and userswith health specialists can be performed remotely [1 2]decreasing overall costs and increasing quality of lifereducing patients displacement and allowing them to live

in their homes [3 4] Parameters such as biomedical signalsdrug distribution patientsrsquo behavior interactions and alarmsignals can be readily collected and analyzedThe interactionof these localized AAL solutions within a smart city or smartregion gave rise to the Smart Health concept [5]

The implementation of context-aware environmentsrelies on the use of a wide variety of communication systemsand very particularly wireless communication solutionswhich allow seamless connectivity by means of differentwireless infrastructures [6] In this sense sustained use ofwireless communication systems has led to the adoption ofadaptive modulation and coding and spectrum allocationschemes in order to optimize coveragecapacity require-ments In this sense interference control plays a vital rolein the performance of wireless systems particularly in 4Gand 5G communication systems Overall power spectraldensity of interference depends on network topology spatial

Hindawi Publishing CorporationJournal of SensorsVolume 2016 Article ID 2424101 16 pageshttpdxdoiorg10115520162424101

2 Journal of Sensors

concentration of users and intrinsic characteristics of net-work terminals and access pointsbase stations When con-sidering the implementation ofAAL environments and SmartHealth solutions we need to face scenarios that are complexin terms of wireless propagation due to the presence ofmultiple elements such as furniture building structure orpeople which can give rise to strong fading effects in anonuniform manner Moreover user density also affectsinterference levels with density values that can exhibit strongvariations as a function of scenario location or time ofanalysis particularly in the case of interaction with wearabledevices wireless body area networks or device-to-deviceconnections Accurate estimations of these interference levelswill help to avoid possible communication errors which insome cases like some e-Health applications where medicalsensors take part could be critical [7]

Wireless channel characterization in terms of usefulreceived signal levels for a given set of connections as wellas for potential interfering connections can be a challeng-ing task due to the large size of scenarios the existenceof multiple frequency dependent materials and inherentvariability of mobile connections given by the movementof potential scatterers In this sense several approachescan be followed from empirical estimations which pro-vide results at low computational cost with low precision(requiring site-specific calibration measurements) to full-wave simulation techniques which provide high accuracy atvery large computational cost As a midpoint deterministictechniques such as ray launching can provide adequateaccuracy while reducing computational cost As the scenariosize grows ray launching techniques increase their com-putational cost In this sense approaches such as sourcedefinition based on Huygens box approximation or the com-bination of 3D ray launching with neural networks reducethe computational effort when the size of the scenario grows[8 9]

In this paper novel approximation based on the com-bination of an in-house developed 3D ray launching codeand a collaborative filtering (CF) technique of bidimensionalcalculation points is used to analyze the performance ofwireless channels emulating AAL scenarios The studiedscenarios have an inherent complexity and a large numberof users and interferers The aim of the proposed method isto provide optimal deployment strategies formassive wirelesssystem and wireless sensor networks The application ofthe presented method in combination with new optimizingalgorithms [10ndash12] will enhance the WSNs performance inAAL and context-aware environments

2 Hybrid Simulation Technique

In this section the description of the in-house developed 3Dray launching code combined with a collaborative filteringapproach is presented This hybrid method will be used toanalyze the behavior of ZigBee wireless sensor networks(WSNs) in AAL environments with the aim of drasticallyreducing the computational time required to obtain accuratesimulation results

21 3D Ray Launching The 3D ray launching code is basedon Geometrical Optics (GO) and the Uniform Theory ofDiffraction (UTD)The principle of the algorithm is that raysare launched in a solid angle with input angular resolutionof parameters 120579 and Φ Electromagnetic phenomena such asreflection refraction and diffraction are taken into accountWe also consider the properties of all the obstacles within theenvironment considering the conductivity and permittivityof their different materials at the frequency of the systemunder analysis It is worth noticing that a grid is definedin the space and the ray launching parameters are storedat each cuboid during the propagation of each ray Theconfiguration parameters of the system are frequency powergain polarization and directivity of the transceivers bit rateangular resolution of the launching and diffracted rays andnumber of reflections The received power is calculated withthe sum of the electric field vectors inside each cuboid of thedefined mesh When a ray hits an obstacle new reflected andtransmitted rays are generated with new angles provided bySnellrsquos lawWhen a ray hits an edge a new family of diffractedrays is generated as we can see in Figure 1 with the diffractioncoefficients given by the UTD which are shown in (1) givenby [13 14]

119863perp=minus119890(minus1198951205874)

2119899radic2120587119896cot119892(

120587 + (Φ2 minus Φ1)

2119899)

sdot 119865 (119896119871119886+(Φ2 minus Φ1)) + cot119892(

120587 minus (Φ2 minus Φ1)

2119899)

sdot 119865 (119896119871119886minus(Φ2 minus Φ1)) + 119877

perp

0

sdot cot119892(120587 minus (Φ2 + Φ1)

2119899)119865 (119896119871119886

minus(Φ2 + Φ1))

+ 119877perp

119899 cot119892(120587 + (Φ2 + Φ1)

2119899)

sdot 119865 (119896119871119886+(Φ2 + Φ1))

(1)

where 119899120587 is the wedge angle 119865 119871 and 119886∓ are defined in [13]and 1198770119899 are the reflection coefficients for the appropriatepolarization for the 0 face or 119899 face respectivelyThe completeapproach has been explained in detail in [15] and it has beenused successfully in different complex indoor environments[16ndash22]

Two simulation configurations have been used to fit theCF method with results High Definition (HD) simulationsand Low Definition (LD) simulations HD simulations useparameters usually set to obtain accurate estimations inindoor environments On the contrary LD simulations pro-vide less accurate estimations but the computational cost ismuch lower The main parameters that have been used forboth LD and HD simulations are summarized in Table 1

22 Recommender Systems and Collaborative Filtering Rec-ommender Systems (RS) are a family of techniques used tomanage and understand the information created by users in

Journal of Sensors 3

Diffraction

Refraction

Reflection

ReRefrac

Figure 1 Simplified representation of the operation principle of the in-house 3D ray launching algorithm

Table 1 Parameters for the 3D ray launching simulations

LD HDFrequency 24GHz 24GHZZigBee transmittedpower 0 dBm 0 dBm

WiFi transmitted power 20 dBm 20 dBmAntenna gain 15 dBi 15 dBiDiffraction enabled No YesHorizontal plane angleresolution (ΔΦ) 2∘ 1∘

Vertical plane angleresolution (Δ120579) 2∘ 1∘

Maximum permittedreflections 3 7

Cuboids resolution 10 cm times 10 cmtimes 10 cm

10 cm times 10 cmtimes 10 cm

Web 20 websites and to allow them to obtain recommen-dations about products and services [23] RS help users todistinguish between noise and profitable information henceachieving their goals more efficiently Moreover thanks toRS companies increase their revenue reduce their costs andprovide better services to their customers

In this paper we use a special kind of RS called collab-orative filtering [24] The aim of CF is to make suggestionson a set of items (119868) (eg books music films or routes)based on the preferences of a set of users (119880) that have alreadyacquired andor rated some of those items In order to makerecommendations (ie to predict whether an item wouldplease a given user) CF methods rely on large databases withinformation on the relationships between sets of users anditems

These data take the form of matrices composed of 119899 usersand119898 items and eachmatrix cell (119894 119895) stores the evaluation of

user 119894 on item 119895 Recommendations provided by CF methodsmake the assumption that similar users are interested inthe same items Hence 119906119886rsquos high rated items could berecommended to user 119906119887 if 119906119886 and 119906119887 are similar CFmethodsare classified into three main categories according to thedata they use (i) memory-based methods which use thedata matrix with all entries ratings and relationships (ii)model-based methods which estimate statistical models andfunctions based on the data matrix and (iii) hybrid methodswhich combine the previous methods with content-basedrecommendation [25]

In this paper we use a memory-based approach to finelytune (recommend) the results obtained by LD simulationsand produce better values that resemble HD simulations inmuch less time Our CF approach is divided into two stepsneighborhood search and recommendationprediction compu-tation In the neighborhood search phase given a user 119906119886 isin119880 we use similarity functions to determine the users thatare most similar to 119906119886 (ie the neighborhood of 119906119886) In therecommendationprediction computationphase we determinethe neighborhood of 119906119886 and make a recommendation usingwell-known methods such as the ones proposed in [25] Theinterested reader could refer to [26ndash29] to delve intoCFrsquos stateof the art

23 Applying Collaborative Filtering on 3D Ray LaunchingSimulations We represent simulation scenarios obtained by3D ray launching techniques as matrices with dimensions119899 times 119898 Without loss of generality we represent each matrix119872119899times119898 as a row vector 119881 that results from the concatenationof all rows of119872 Each scenario may contain different sets ofobstacles and materials Our goal is to predict (recommend)values that LD simulations were unable to compute properlydue to their low resolution To do so we follow the idea


V = V1 V2 V3 V4 middot middot middot

V2 V3

V1 V2 V3

V4

middot middot middot

SV1 =

SV2 =

SV(Lminus2) =

minusLSV + 1

VL119881minus2VL119881minus1

VL119881


VL119881

Figure 2 Graphical scheme of subvectors generation

proposed in [30]where amemory-basedCFmethodwith twostages was suggestedThose stages are (i) knowledge databasecreation and (ii) values prediction

Knowledge Database Creation This stage consists in thecreation of a database that will be later used to predictmissingvalues in LD simulations (in the second stage) For eachscenario represented by a vector 119881 = (V1 V2 V119871119881) withlength 119871119881 we create a collection of subvectors with fixedlength 119871SV SV = sv1 sv2 sv(119871119881minus119871SV+1) so that sv119894 =(V119894 V119894+1 V119871SV+119894minus1) forall119894 | 119871119881minus119871SV+119894 ge 0 Figure 2 shows howsubvectors with 119871SV = 3 are generated from a given vectorscenario 119881 This process results in a database comprisingfixed-length subvectors that are used as patterns representinga variety of scenarios Note that we can create severaldatabases with different subvector lengths containing asmany scenarios as we want Also it is worth noticing that thisprocedure is applied to vectors of LD simulations and theircorresponding HD counterparts As a result the databasescontain LD patternsvectors that are correlatedassociatedwith their corresponding HD counterparts

With the aim of increasing the quality of the databasewe select LD subvectors that have similarhighly correlatedvalues to their HD counterparts in this way we reduce thematching error between LD and HD in the database Todo so we fix a maximum distance threshold between LDand HD values for each subvector and we discard those thatsurpass this value (for each LD simulation in the knowledgedatabase we have its corresponding HD simulation andwe can compare their results) Moreover we have observedthat LD measurements in the vicinity of cells that have notbeen properly simulated (eg due to the simulation lack ofaccuracy in LD) and have null or error values do not correlatewell with their HD counterparts even when they fall withinthe aforementioned threshold Hence in order to solve thisproblem we discard values in the vicinity of those cells Todo so we compute the Manhattan distance between the cellscontaining nullerror values and the others Next we set aminimum distance (ie in our case 3 due to the minimumsubvector length) andwe discard those LD valuescells whosedistance is lower So in essence we discard all cells that arelocated atManhattan distances equal to or lower than 2 fromanullerror cell Figure 3 shows an example of howManhattandistances are computed After applying these noise-reducingtechniques the result is a more reliable and robust database

55

5 5

6 6 4

4

4 4

4

43

3

3 3

3

3

3

3

3

33

2

2

2

2

2

2

2

2

2

2

2

2

2

1

1

1

1

1

1

1

1

1

minus1

minus1

minus1

minus1

Figure 3 Example of theManhattan distance computation betweennull cells (minus1 highlighted in red) and the others Considering athreshold distance = 3 possible noisy cells are detected (highlightedin light orange) Any subvector containing ldquolight orangerdquo cells isdiscarded

Subvector with missing value

V1 V2 V4 V5

VPV1 V2 V4 V5

(1) Find subvectorrsquos k closest patterns

(2) Corresponding HD values of k

LD database HD database

(3) Aggregate HD patterns and obtain predicted value VP

Figure 4 (1)We find 119896 subvectors with the closest patterns in the LDdatabase (2)We compute the average of their associated HD values(3) Finally we replace the missing value by the HD average

Values Prediction Given an LD simulation 119878 with miss-ingnull values our goal is to predict them so that they areas similar as possible to the values that would have beenobtained in an HD simulation To do so the values of 119878 arenormalized to be comparable with those in the LDknowledgedatabase Next 119878 is divided into subvectors of a chosenlength 119871SV (like in the previous step) For each subvectorcontaining missing values 119896 most similar subvectors inthe LD knowledge database (created in the previous step)are found Then their corresponding HD subvectors areretrieved and the missing values in 119878 are replaced by theaverage of those HD values

To compute the similarity between subvectors we usethe Euclidean distance over nonmissing values A graphicalrepresentation of this procedure is depicted in Figure 4 With


the aimof increasing the predictions quality we only considersubvectors having at most one missing value Moreover foreach vector we compute the percentage of empty cells andavoid computations if this percentage is higher than 50 (itwould be useless to predict any value without information todo so in such a case the choice would be mostly random)thus increasing the performance of our method

The prediction procedure is first applied by rows andnext by columns and then we compute the average of theoutputs (predicted values are only considered if they havebeen predicted both by rows and by columns (otherwisethey are discarded)) Finally we compute the mean of theobtained predictions in order to fill those cells that could notbe inferred using the aforementioned approach It is worthmentioning that the prediction procedure is applied for eachsubvector length Hence we will have different outcomesdepending on 119871SV

3 ZigBee Network Performance Analysis

In this section we describe the methodology followed toanalyze the performance of a dense ZigBee wireless networkin AAL environments and to validate our hybrid methodbased on CF First the description of the scenario underanalysis is presented Then with the aim of validating thesimulation results obtained by the in-house 3D ray launchingalgorithm a measurement campaign was carried out to com-pare real measurements with the estimated values Finallyonce the results obtained by the 3D ray launching methodhave been validated the performance analysis of a denseZigBee network deployed within the scenario is presentedFor that purpose the estimations obtained by both the HDray launching method and the hybrid LD + CF method areshown

31 Description of the Scenario under Analysis The scenarioconsidered in this paper is a common apartment located inthe neighborhood of ldquoLa Milagrosardquo in the city of PamplonaNavarre (Spain) The apartment is approximately 65m2 andit consists of 2 bedrooms 1 kitchen 1 bathroom 1 studyroom 1 living room and 1 small box room as it can be seenin Figure 5 The dimensions of the scenario are 905m times

7255m times 2625m In order to obtain accurate results with the3D ray launching simulation tool the real size and materialproperties (dielectric constant as well as conductivity) of allthe furniture such as chairs tables doors beds wardrobesbath and walls have been taken into account In Table 2the properties of the materials with greater presence in thescenario are listed

Due to the increasing number of wireless systems andapplications for AAL scenarios like the apartment studied inthis paper are expected to need a large number of wirelessdevices deployed in quite small areas In order to emulatea dense wireless network a 56-device ZigBee network hasbeen distributed throughout the apartment Those devicescould be either static ormobile andwearable devices Figure 5shows the distribution of the devices ZigBee End Devices(ZEDs) are represented by red dots and ZigBee Routers (ZR)

Table 2 Properties of the most common materials for our 3D raylaunching simulations

Material Permittivity (120576119903) Conductivity (120590) [Sm]Air 1 0Plywood 288 021Concrete 566 0142Brick wall 411 00364Glass 606 10minus12

Metal 45 4 times 107

Polycarbonate 3 02

Table 3 Parameters of the wireless devices for ray launchingsimulations

Parameter ValueFrequency of operation 24GHzWiFi antenna type MonopoleWiFi access point transmission power 20 dBmWiFi device transmission power 20 dBmWiFi antenna gain 15 dBiZigBee antenna type MonopoleZigBee antenna gain 15 dBiZigBee transmission power 0 dBm

by green dots In order to make the AAL environment morerealistic non-ZigBee wireless devices have been also placedwithin the scenario so as to analyze their coexistence interms of intersystem interferences For that purpose a WiFiaccess point (black dot in Figure 5) and a WiFi device thatcould be a laptop or a smart phone (blue dot in Figure 5)have been placed in the living room and in the study roomrespectively The characteristics of the antennas and devicesused in the scenario such as the transmission power level andthe radiation pattern have been defined as those typical forreal devices Table 3 shows the main parameters used in thesimulations for both ZigBee and WiFi devices

32 Validation of the Ray Launching Simulation Results Oncethe simulation scenario has been defined a measurementcampaign has been carried out in the real scenario in orderto validate the results obtained by the 3D ray launchingalgorithm For that purpose a ZigBee-compliant XBee Promodule connected to a computer through a USB cable hasbeen used as a transmitter (cf Figure 6(a)) The transmissionpower level of the wireless device can be adjusted from0 dBm to 18 dBm For the validation the 0 dBm level hasbeen set which will be the level used for the simulationsof the ZigBee network reported in the following sectionsReceived power values in different points within the sce-nario have been measured by means of a 24GHz centeredomnidirectional antenna connected to a portable N9912AFieldFox spectrum analyzer (cf Figure 6(b)) In order tominimize the spectrum analyzerrsquos measurement time thetransmitter has been configured to operate sending one datapacket per millisecond The measurement points and theposition of the transmitter element depicted in the schematic


7255m

905

m

2625

m(a)

Aisle

Bathroom

Kitchen

Box room

Study room

Living room

Bedroom 2

Bedroom 1

X

Y

ZRZED

WiFi access point WiFi device

(b)

Figure 5 (a) 3D simulated scenario with wireless devices shown in coloured dots (b) top view of the scenario

(a) (b)

Figure 6 (a) XBee Pro module (b) Portable N9912A FieldFox spectrum analyzer used

view shown in Figure 7 are represented by green pointsand a red rectangle respectively The transmitter has beenplaced at height 08m and the measurements have beentaken at height 07m The obtained power levels for eachmeasurement point have been depicted in Figure 8 wherea comparison between measurements and 3D ray launchingsimulation results is shown It may be observed that the 3Dray launching simulation tool provides accurate estimationsIn this case the outcome is a mean error of 138 dB with astandard deviation of 255 dB

33 ZigBee Network Performance Analysis Following thevalidation of the 3D ray launching simulation algorithm thefeasibility and performance analysis of the ZigBee networkdeployed in the scenario are presented A ZigBee-basedwireless network has been deployed within the scenario(cf Figure 5) The network consists of 50 ZEDs whichemulate different kinds of sensors distributed throughout

the apartment and 6 ZR which emulate the devices that willreceive the information transmitted by the ZEDs includedin their own network Finally a WiFi access point has beenplaced in the living room and a WiFi device in the studyroom In the real scenario there is a WiFi access point in thesame placeThese wireless elements which are very commonin real scenarios could interfere with a ZigBee networkIn order to gain insight into this issue two spectrogramshave been measured in the aisle of the real scenario First aspectrogram with only the WiFi access point operating hasbeen measured Then an operating XBee Pro mote has beenmeasured with theWiFi access point off Figure 9 shows bothspectrogramsTheXBee Promodule is operating at the lowestfrequency band allowed for these devices and it can be seenhow its spectrum overlaps the WiFi signal Although WiFiand ZigBee can be configured to operate at other frequencybands the bandwidth of the WiFi signal is wider than theZigBee signalThus sincewemaydeploy lots of ZigBee-based


Transmitter

P14

P13

P12

P10 P9 P8 P7 P6P5

P4P11

P3

P2

P1

1 2 3 4 5 6 7 8 90

X-distance (m)

0

1

2

3

4

5

6

7

Y-d

istan

ce (m

)

Figure 7 Schematic of the scenario Measurement points are shown in green and the transmitter is shown in red

P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 P13 P14Measurement points

MeasurementsSimulation results

minus70minus65minus60minus55minus50minus45minus40minus35minus30

Rece

ived

pow

er (d

Bm)

Measurements versus simulation results

Figure 8 Comparison of real versus 3D ray launching simulationresults for different points (cf Figure 7)

networks each one operating at different frequency bandsit is likely that overlapping between such wireless systemsoccurs

Therefore in addition to the performance of a WSN inAAL environments in terms of received power distributionthe assessment of the nondesired interferences is a majorissue especially in complex indoor environments wheremany wireless networks coexist and where radiopropagationphenomena like diffraction and fast fading are very strongThe effect of these phenomena as well as the topology andmorphology of the scenario under analysis can be stud-ied with the previously presented simulation method Thedeployment of ZigBee-based WSNs is versatile and allowsseveral topology configurations such as star mesh or tree

However due to the wireless inherent properties and the sizeof the motes their position and the network topology itselfcan be easily modified and reconfigured Hence very differ-ent networks and subnetworks may be discovered in a singleAAL environmentThe 3D ray launching algorithm is an ade-quate tool to analyze the performance of this kind of wirelessnetwork as it has been shown in previous sectionsThe mainresult provided by the simulations is the received power levelfor the whole volume of the scenario Figure 10 shows anexample of the received power distribution in a plane at aheight of 15m for the transmitting ZED placed within thebox room It can be observed that the morphology of thescenario has a great impact on the results In this exampleit is shown how the radiopropagation through the box roomrsquosgate and through the isle is greater than for further roomsbecause there are less obstacles and walls in the rays path

The main propagation phenomenon in indoor scenarioswith topological complexity is multipath propagation whichappears due to the strong presence of diffractions reflectionsand refractions of the transmitted radio signals Multipathpropagation typically produces short-term signal strengthvariations This behavior can be observed in Figure 11where the estimated received power versus linear distanceis depicted for 3 different heights corresponding to thewhite dashed line of Figure 10 The relevance of multipathpropagation within the scenario can be determined by theestimation of the values of all the components reached at aspecific point usually the receiver In order to assess thisPower Delay Profiles are utilized Figure 12 shows the PowerDelay Profiles for 3 different positions within the scenariowhen the ZED of the box room is emitting As it may be


2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)

minus60dBmminus55dBmminus50dBm

minus45dBmminus40dBm

(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)

Figure 9 Spectrograms in the real scenario (a) WiFi access point only (b) XBee Pro module only

minus100dBmminus90dBmminus80dBmminus70dBmminus60dBm

minus50dBmminus40dBmminus30dBmminus20dBm

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED

Figure 10 Received power distribution plane at a height of 15mwhen the ZED within the box room is transmitting The whitedashed line corresponds to the data shown in Figure 11

observed the complex environment creates lots of multipathcomponents which are very strong in the box room as thetransmitter is in it As expected these components are fewerand weaker (ie lower power level) as the distance to thetransmitter grows as shown in the Power Delay Profiles ofrandomly chosen points within the aisle and the living roomNote that the arrival time (x-axis) of the first component ofeach graph clearly shows the increasing distance being higherwhile the distance growsThe dashed red lines depicted in thePowerDelay Profiles correspond to the sensitivity of the XBee

1m height15m height2m height

minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)

Figure 11 Estimated received power versus linear distance for 3different heights corresponding to the white dashed line depictedin Figure 10

Pro modules indicating which of the received componentscan be read by the receiver those with a power level higherthan minus100 dBm An alternative approach to show the impactof multipath propagation is the use of Delay Spread graphswhich provide information for a whole plane of the scenarioin a single graph instead of the ldquopoint-likerdquo informationprovided by Power Delay Profiles The Delay Spread is thetimespan from the arrival of the first ray to the arrival of thelast ray at any position of the scenario that is the timespanbetween the first and the last element depicted in PowerDelayProfile figures Figure 13 shows the Delay Spread for a planeat a height of 2m when the ZED placed in the box room


Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)

Figure 12 Estimated Power Delay Profiles when the ZigBee mote within the box room is emitting for different points within the scenario(a) box room (b) aisle and (c) living room

transmits As expected larger timespan values can be foundin the vicinity of the emitting ZED since this is the area wherethe reflected valid rays (ie rays with power level higherthan the sensitivity value) arrive sooner In other areas DelaySpread values vary depending on the topology of the scenario

The Power Delay Profiles and Delay Spread results showthe complexity of this kind of scenario in terms of radioprop-agation and the amount of multipath components Assumingthat the number of wireless devices deployed in anAALWSNcan be high radioplanning becomes essential to optimizethe performance of WSNs in terms of achievable data ratecoexistence with other wireless systems and overall energyconsumption One of the most WSN performance limitingfactors is given by the receivers sensitivity which is basicallydetermined by the hardware itself Other limiting factors arethe modulation and the coding schemes which determinethe maximum interference levels tolerated for a successfulcommunicationThe interference level in AAL environmentswith dense WSNs deployed within could be very highnegatively affecting the communication between wirelessdevices measured in terms of increasing packet losses As

an illustrative example the interference and performanceassessment on the ZR deployed in the kitchen (see Figure 5for the position) is shown in Figure 14 Typically the ZEDnodes of the network do not transmit simultaneously as theyare programmed to send information (eg temperature) inspecific intervals or when an event occurs (eg detection ofsmoke) Thus in this example a ZED located in the kitchen(119883 = 49m 119884 = 36m and 119885 = 1m) sends informationto the ZR Figure 14 shows the received power plane at24m height where the ZR is deployed Note that the ZEDis placed at 1m height The power level received by the ZRis minus5026 dBm Considering that the sensitivity of the XBeePro modules is minus100 dBm the received power level is enoughto have a good communication channel Nevertheless thefeasibility of a good communication between both the ZEDand the ZR will depend mainly on the interference levelreceived by the ZR device

The undesirable interference received by a wirelessreceiver could have different sources such as electric appli-ances that generate electromagnetic noise However interfer-ence is mainly generated by other wireless communication


020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m

Figure 13 Delay Spread at 2m height when ZED transmits in thebox room (119883 = 05m 119884 = 28m and 119885 = 1m)



ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR

Figure 14 Received power at 24m height when ZED transmits inthe kitchen (119883 = 49m 119884 = 36m and 119885 = 1m)

systems and devices With the aim of showing the usefulnessof the presented method for the analysis of interferences weanalyze 3 different situations for the wireless communicationbetween a ZED and a ZR deployed in the scenario underanalysis We will study an intrasystem interference case aZigBee intersystem case and a WiFi intersystem case Theinterference level (in dBm) produced by the correspondentsources of the 3 case studies is respectively shown in Figures15 16 and 17 Note that in all cases the interference willoccur only when the valid ZED and the interference sourcestransmit simultaneously and the frequency bands overlap

In order to show a graphic comparison between the HDresults obtained by the 3D ray launching algorithm and ourhybrid approach based LD + CF estimations both of themare shown in each figure For the intrasystem interference

case a ZED within the same network of the valid ZED-ZR devices has been chosen to act as interference source(cf Figure 15) For the intersystem interference analysis twodifferent configurations have been studied On the one hand4 ZigBee modules of another network deployed within thescenario act as interference sources In Figure 16 we show theinterference level when the 4 ZEDs (at 1m height) transmitsimultaneously On the other hand 2 WiFi devices placed at119883 = 7 119884 = 5 and119885 = 08 and119883 = 05 119884 = 13 and119885 = 11act as interference sources It can be observed in Figure 17 thatthe interference level is significantly higher than in previouscases This is due to the transmission power of the WiFidevices which has been set to 20 dBm (ie the maximumvalue for 80211 bgn 24GHz) while that of ZED has beenset to 0 dBm In all cases we confirm that the morphologyof the scenario has a great impact on the interference levelthat will reach the ZRThe number of interfering devices andtheir transmission power level will also be key issues in termsof received interference power Therefore an adequate andexhaustive radioplanning analysis is fundamental to obtainoptimized WSN deployments reducing to the minimum thedensity of wireless nodes and the overall power consumptionof the network For that purpose the method presented inthis paper is a very useful tool

Based on the previous results SNR (Signal-to-NoiseRatio) maps can be obtained The SNR maps provide veryuseful information about how the interfering sources affectthe communication between two elements making it easierto identify zones where the SNR is higher and hence betterto deploy a potential receiver Figure 18 shows SNR mapsfor the 3 interference cases previously analyzed For someapplications it might not be possible to choose the receiverrsquosposition nor the position of other wireless emitters In thosecases instead of using SNRmaps estimations of SNR level atthe specific point where the receiver is placed are calculatedFigure 19 shows the SNR value at the position of the ZR forthe previously studied cases Note that the x-axis representspossible transmission power levels of the valid ZED Againthe estimations by the HD ray launching as well as estimatedvalues by LD ray launching + collaborative filtering (LD +CF) are shown in order to show the similarity between bothmethods in terms of predicted values The red dashed linesin the figure represent the minimum required SNR value fora correct transmission between the valid ZED and the ZR at256Kbps and 32Kbps which have been calculated with theaid of the following well-known formula

119862 = BW log2 (1 +119878

119873) (2)

where 119862 is the channel capacity (250Kbps maximum valuefor ZigBee devices and 32Kbps) and BW is the bandwidth ofthe channel (3MHz for ZigBee) As it can be seen in Figure 19for the analyzed ZigBee intersystem interference case (greencurve) estimations show that it is likely to have no problemand the transmission will be done at the highest data rate(256Kbps) even for a transmission power level of minus10 dBmFor the intrasystem case (black curve) a communicationcould fail when the transmitted power decays to minus10 dBmwhich means that the SNR value is not enough to transmit




ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)

Figure 15 Received intrasystem interference level at 24m height when a second ZED in the kitchen (119883 = 47m 119884 = 63m and 119885 = 1m) istransmitting (a) HD results and (b) LD + CF estimations

ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)

Figure 16 Received ZigBee intersystem interference level at 24m height when four ZEDs belonging to a network in bedroom 1 aretransmitting (a) HD results and (b) LD + CF estimations

at 256Kbps but it is still enough to successfully achievelower data rate transmissions Finally under the conditionsof the 2 WiFi devices interfering with the ZED-ZR linkthe communication viability depends strongly on the trans-mission power level of the valid ZED even for quite lowdata rates as the noise level produced by the WiFi devicestransmitting at 20 dBm is high It is important to note thatthese SNR results have been calculated for a specific case with

specific transmission power levels for the involved devicesDue to the huge number of possible combinations of theprevious factors (number position and transmission powerof interfering sources valid communicating devices requireddata rates morphology of the scenario etc) the designprocedure is site specific and thus the estimated SNR willvary if we change the configuration of the scenariorsquos wirelessnetworks Therefore the presented simulation method can


WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)

Figure 17 Received WiFi interference level at 24m height (a) HD results and (b) LD + CF estimations

help in estimating the SNRvalues at each point of the scenariofor anyWSNconfiguration allowing the designer tomake thecorrect decision in order to deploy and configure the wirelessdevices in an energy-efficient way

4 LD + CF Prediction Quality Analysis

In previous sections we have described our hybrid LD +CF proposal and we have shown that the results obtainedwith this method are similar to those obtained with an HDsimulation Next we summarize how these results have beenobtained and we evaluate in detail the difference betweenthe computational costs of HD simulations and the LD + CFapproach

41 Knowledge Database Creation To create the knowledgedatabases used by the LD +CFmethod 16 different scenarioshave been simulated with 30 to 40 layers each containinga variety of features (ie corridors columns walls doorsand furniture) Each scenario has been simulated in LDand HD With these simulations we have created five LDknowledge databases with 119871SV = 11 9 7 5 and 3 and theirfive HD counterparts Each knowledge database containsapproximately one million patternssubvectors In this studythe scenario under analysis has similar characteristics tothose which have been used for the creation of the databaseIts dimensions and density are summarized in Table 4 Rowsand Columns refer to the planar dimensions of the scenario(ie the dimensions (119883119884) of the matrices analyzedusedby the CF method) The Layers row indicates the numberof matrixes that is the height (119885) of the scenario Finallythe Density row shows the percentage of the volume of thescenario occupied by obstacles

Table 4 Dimensions and characteristics of the scenario

Rows Columns Layers Density Scenario 91 73 27 9497

Table 5 Simulation strategies subvector length 119871SV and aggregatorvalue 119896

Strategy Details1 119871SV = 11 119896 = 252 119871SV = 9 119896 = 253 119871SV = 7 119896 = 254 119871SV = 5 119896 = 255 119871SV = 3 119896 = 25

42 Accuracy and Benefits of the Collaborative FilteringApproach With the aim of analyzing the accuracy andperformance of our approach different prediction strategieshave been applied for the scenario under analysis (cf Table 5)For instance Strategy 1 uses the previously created knowledgedatabase with 119871SV = 11 to compute predictions in Strategy2 we use 119871SV = 9 and so on In all cases an aggregatorvalue 119896 = 25 is used Hence for each missing value ofan LD simulation the CF approach finds 119896 = 25 mostsimilar subvectors and computes their average to predict themissing value Although the CF approach significantly helpsto improve the quality of the LD simulation it does not alwayspredict the exact same value of the HD simulation In orderto compute this discrepancy we use the well-known meanabsolute error ldquoMAErdquo defined in (3) as follows

MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

minus50dBminus40dBminus30dBminus20dB

minus10dB0dB10dB20dB

(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)

Figure 18 SNRmaps for the 3 different interference sources acting over the valid ZED (a) intrasystem (b) ZigBee intersystem and (c)WiFi

where 119899 is the number of missing values predicted 119901119894 is thepredicted value for a missing element 119894 and 119903119894 is the real valueof 119894 in the HD simulation Note that the HD simulation isonly used to compute the error but it is not involved in theprediction process of the CF method

Without loss of generality and for the sake of brevity wehave randomly selected 24 sensors from the studied scenarioandwe have compared the simulation results obtained byHDsimulations and our hybrid approach LD+CF Table 6 reportsthe obtained results Specifically Table 6 reports Sparsenessthat is the of empty values in the LD simulation Time HDTime LD Time CF and Time LD + CF which are the time inseconds of each method (for LD + CF we report the worsttime not the average) Best Strategy which is the strategy that

performed better MAE119877 which represents the mean absoluteerror considering that null cells are kept empty (ie LDversus HD) MAE119875 which represents the mean absolute errorconsidering that null cells are replaced by the values predictedby the CF method (ie LD + CF versus HD) and 120590119875 that isthe standard deviation of MAE119875

It is worth noting the difference between MAE119877 andMAE119875 Note that for the mean absolute error the lowerthe values the better the result Hence it is apparent thatthe results obtained by our hybrid method (MAE119875) clearlyoutperform those of the LD simulation alone (MAE119877)Moreover this significant improvement requires very littletime and in addition the low values of 120590119875 indicate thatpredictions are stable and reliable It may also be observed


Table 6 Average results (over all layers of the scenario) of the hybrid LD + CF versus the HD approach

Sparseness Time HD Time LD Time CF Time LD + CF Best Strategy MAE119877 MAE119875 120590119875

Sim1 47221 90650 2981 112 3116 1 74707 11825 905Sim2 49933 52309 3543 187 2783 4 75181 13252 1022Sim3 30291 64959 2953 55 3302 1 73091 9796 747Sim4 49071 53709 3168 106 2948 1 72737 10812 836Sim5 41458 58763 3197 116 3053 3 75051 1052 856Sim6 42722 83949 3004 88 2730 1 72509 10281 757Sim7 33574 90189 2596 84 3362 1 72439 988 712Sim8 34711 76368 3247 102 3211 2 73408 9914 749Sim9 39952 94287 2842 103 3153 1 74996 9814 782Sim10 2003 69627 2937 79 2889 2 78893 7797 625Sim11 24915 80113 3315 99 2947 3 7838 972 767Sim12 35317 79136 2882 98 3246 1 70076 13394 892Sim13 35266 56627 2608 101 2982 1 72139 12021 883Sim14 43428 59785 2906 115 2776 1 73255 14092 101Sim15 43154 46410 2870 138 2558 4 76506 12005 911Sim16 38066 67031 2642 95 2971 1 7224 13756 994Sim17 45551 50729 3278 123 2509 2 76485 13294 989Sim18 42902 83750 3109 120 3106 1 69985 14618 1073Sim19 55942 50423 3050 168 2761 2 74522 12073 891Sim20 54066 58216 2810 186 3109 3 72824 11724 88Sim21 57699 55179 3183 201 2985 3 74018 10716 829Sim22 44755 74635 2606 99 2869 1 71712 14338 1088Sim23 38979 59727 2877 56 2983 1 64586 16998 1145Sim24 48459 75307 3245 79 2794 1 74443 10435 786

minus25minus20minus15minus10

minus505

10152025

minus10 0 10 18

SNR

(dB)

Transmitted power (dBm)

Intrasystem HD Intrasystem LD + CFZigBee intersystem HD ZigBee intersystem LD + CFWiFi HD WiFi LD + CF

256Kbps

32Kbps

Figure 19 Estimated SNR values at ZR for 3 different interferencecases

that our hybrid approach requires 10 to 20 times less timethan HD simulations and that the cost of the CF method isalmost negligible (cf Table 6 column ldquoTime LD + CFrdquo andFigure 20) Also we observe that sparseness has an adverseeffect on the prediction accuracy of all methodsThis result isnot surprising since with less data it is more difficult to makebetter decisions (however the knowledge database is also animportant factor as can be observed in Sim21 and Sim23

where Sim21 has a higher sparseness value but the predictionrsquosaccuracy (ie MAE119875) is far better than in Sim23 whichexhibits the opposite behavior) Since the best prediction isnot always obtained by the same CF strategy if more accurateresults were to be obtained parameters such as 119896 subvectorlength and its corresponding knowledge database might betuned depending on each simulationrsquos features

The relationship between the MAE of each approachand simulationprediction time is depicted in Figure 21 LDpredictions (in red) correspond to MAE119877 values in Table 6LD + CF results (in green) show that the computationaltime is slightly increased with respect to LD simulationsNotwithstanding the MAE is reduced between 8 and 10times Finally the values of HD simulations (in blue) clearlyshow that the time required is 10 to 20 times higher than theother approaches Clearly the best tradeoff betweenMAEandtime is obtained by our proposed hybrid LD + CF methodHence we may conclude that our method outperforms theothers when we consider both accuracy and computationalcost

5 Conclusions

In this paper context-aware scenarios for ambient assistedliving have been analyzed in the framework of the deploy-ment of wireless communication systems (mainly wearabletransceivers and wireless sensor networks) and the impact oncoveragecapacity relations


Times LD + CFTimes LDTimes HD

Sim1Sim2Sim3Sim4Sim5Sim6Sim7Sim8Sim9

Sim10Sim11Sim12Sim13Sim14Sim15Sim16Sim17Sim18Sim19Sim20Sim21Sim22Sim23Sim24

Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)

Figure 20 Time comparison of the different approaches (ie LDLD + CF and HD) for each simulation

HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)

Figure 21 Relationship between MAE and execution time for LDLD + CF and HD approaches

We have presented and discussed the use of ray launchingsimulations in High Definition (HD) and Low Definition(LD) and collaborative filtering (CF) techniques A complexindoor scenario with multiple transceiver elements has beenanalyzed with those techniques and the obtained results showthat the proposed hybrid LD + CF approach outperforms LDand HD approaches in terms of errortime ratio

We have shown that radiopropagation in indoor complexAAL environments has strong dependence on the networktopology the indoor scenario configuration and the densityof usersdeviceswithin it Results also show that the presentedhybrid calculation approach enables us to enlarge the sce-nario size without increasing computational complexity or

in other words to reduce drastically the simulation time con-sumption while the error of the estimations remains lowOurproposal provides valuable insight into the network designphases of complex wireless systems typical in AAL whichderives in optimal network deployment reducing overallinterference levels and increasing overall systemperformancein terms of cost reduction transmission rates and energyefficiency

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors wish to acknowledge the support providedunder Grant TEC2013-45585-C2-1-R funded by theMinistryof Economy and Competitiveness Spain URV authors arepartly supported by La Caixa Foundation through ProjectSIMPATIC and the Spanish Ministry of Economy and Com-petitiveness under Project ldquoCo-Privacy TIN2011-27076-C03-01rdquo and by the Government of Catalonia under Grant 2014-SGR-537

References

[1] F Palumbo P Barsocchi F Furfari and E Ferro ldquoAALmiddle-ware infrastructure for green bed activity monitoringrdquo Journalof Sensors vol 2013 Article ID 510126 15 pages 2013

[2] A Leone G Diraco and P Siciliano ldquoContext-aware AAL ser-vices through a 3D sensor-based platformrdquo Journal of Sensorsvol 2013 Article ID 792978 10 pages 2013

[3] S Led L Azpilicueta E Aguirre M M D Espronceda LSerrano and F Falcone ldquoAnalysis and description of HOLTINservice provision for AECG monitoring in complex indoorenvironmentsrdquo Sensors vol 13 no 4 pp 4947ndash4960 2013

[4] A Moreno I Angulo A Perallos et al ldquoIVAN intelligent vanfor the distribution of pharmaceutical drugsrdquo Sensors vol 12no 5 pp 6587ndash6609 2012

[5] A Solanas C Patsakis M Conti et al ldquoSmart health a context-aware health paradigm within smart citiesrdquo IEEE Communica-tions Magazine vol 52 no 8 pp 74ndash81 2014

[6] E Aguirre M Flores L Azpilicueta et al ldquoImplementing con-text aware scenarios to enable smart health in complex urbanenvironmentsrdquo in Proceedings of the IEEE International Sympo-sium on Medical Measurements and Applications (MeMeA rsquo14)pp 508ndash511 Lisboa Portugal June 2014

[7] D Lin X Wu F Labeau and A Vasilakos ldquoInternet of vehiclesfor E-health applications in view of EMI on medical sensorsrdquoJournal of Sensors vol 2015 Article ID 315948 10 pages 2015

[8] V Degli-Esposti F Fuschini E M Vitucci and G FalciaseccaldquoSpeed-up techniques for ray tracing field prediction modelsrdquoIEEE Transactions on Antennas and Propagation vol 57 no 5pp 1469ndash1480 2009

[9] L Azpilicueta M Rawat K Rawat F M Ghannouchi and FFalcone ldquoA ray launching-neural network approach for radiowave propagation analysis in complex indoor environmentsrdquoIEEE Transactions on Antennas and Propagation vol 62 no 5pp 2777ndash2786 2014


[10] A Ahmad M M Rathore A Paul and B-W Chen ldquoDatatransmission scheme using mobile sink in static wireless sensornetworkrdquo Journal of Sensors vol 2015 Article ID 279304 8pages 2015

[11] N G S Campos D G Gomes F C Delicato A J V NetoL Pirmez and J N de Souza ldquoAutonomic context-awarewireless sensor networksrdquo Journal of Sensors vol 2015 ArticleID 621326 14 pages 2015

[12] T K Jain D S Saini and S V Bhooshan ldquoLifetime optimiza-tion of a multiple sink wireless sensor network through energybalancingrdquo Journal of Sensors vol 2015 Article ID 921250 6pages 2015

[13] R J Luebbers ldquoA heuristic UTD slope diffraction coefficientfor rough lossy wedgesrdquo IEEE Transactions on Antennas andPropagation vol 37 no 2 pp 206ndash211 1989

[14] R J Luebbers ldquoComparison of lossy wedge diffraction coeffi-cients with application to mixed path propagation loss predic-tionrdquo IEEE Transactions on Antennas and Propagation vol 36no 7 pp 1031ndash1034 1988

[15] L Azpilicueta M Rawat K Rawat F Ghannouchi and FFalcone ldquoConvergence analysis in deterministic 3D ray launch-ing radio channel estimation in complex environmentsrdquo ACESJournal vol 29 no 4 pp 256ndash271 2014

[16] I Salaberria A Perallos L Azpilicueta et al ldquoUbiquitousconnected train based on train-to-ground and intra-wagoncommunications capable of providing on trip customized digi-tal services for passengersrdquo Sensors vol 14 no 5 pp 8003ndash80252014

[17] L Azpilicueta F Falcone J J Astrain et al ldquoAnalysis of topo-logical impact on wireless channel performance of intelligentstreet lighting systemrdquo Radioengineering vol 23 no 1 pp 412ndash420 2014

[18] A Bermejo J Villadangos J J Astrain et al ldquoOntology basedroad traffic managementrdquo Ad Hoc amp Sensor Wireless Networksvol 1-2 pp 47ndash69 2014

[19] E Aguirre P L Iturri L Azpilicueta et al ldquoAnalysis ofestimation of electromagnetic dosimetric values from non-ionizing radiofrequency fields in conventional road vehicleenvironmentsrdquo Electromagnetic Biology and Medicine vol 34no 1 pp 19ndash28 2015

[20] E Rajo-Iglesias E Aguirre P Lopez L Azpilicueta J Arponand F Falcone ldquoCharacterization and consideration of topo-logical impact of wireless propagation in a commercial aircraftenvironment [wireless corner]rdquo IEEE Antennas and Propaga-tion Magazine vol 55 no 6 pp 240ndash258 2013

[21] I Angulo A Perallos L Azpilicueta et al ldquoTowards a trace-ability system based on RFID technology to check the contentof pallets within electronic devices supply chainrdquo InternationalJournal of Antennas and Propagation vol 2013 Article ID263218 9 pages 2013

[22] L Azpilicueta F Falcone J J Astrain et al ldquoMeasurement andmodeling of a UHF-RFID system in a metallic closed vehiclerdquoMicrowave and Optical Technology Letters vol 54 no 9 pp2126ndash2130 2012

[23] P Resnick and H R Varian ldquoRecommender systemsrdquo Commu-nications of the ACM vol 40 no 3 pp 56ndash58 1997

[24] D Goldberg D Nichols B M Oki and D Terry ldquoUsingcollaborative filtering to weave an information tapestryrdquo Com-munications of the ACM vol 35 no 12 pp 61ndash70 1992

[25] X Su andTMKhoshgoftaar ldquoA survey of collaborative filteringtechniquesrdquoAdvances in Artificial Intelligence vol 2009 ArticleID 421425 19 pages 2009

[26] F Casino C Patsakis D Puig and A Solanas ldquoOn privacy pre-serving collaborative filtering current trends open problemsand new issuesrdquo in Proceedings of the IEEE 10th InternationalConference on e-Business Engineering (ICEBE rsquo13) pp 244ndash249Coventry UK September 2013

[27] F Casino J Domingo-Ferrer C Patsakis D Puig and ASolanas ldquoPrivacy preserving collaborative filtering with k-anonymity through microaggregationrdquo in Proceedings of the10th International Conference on e-Business Engineering (ICEBErsquo13) pp 490ndash497 IEEE September 2013

[28] F Casino J Domingo-Ferrer C Patsakis D Puig and ASolanas ldquoA k-anonymous approach to privacy preserving col-laborative filteringrdquo Journal of Computer and System Sciencesvol 81 no 6 pp 1000ndash1011 2015

[29] Y Shi M Larson and A Hanjalic ldquoCollaborative filteringbeyond the user-itemmatrix a survey of the state of the art andfuture challengesrdquoACMComputing Surveys (CSUR) vol 47 no1 pp 1ndash45 2014

[30] F Casino L Azpilicueta P Lopez-Iturri E Aguirre F FalconeandA Solanas ldquoHybrid-based optimization of wireless channelcharacterization for health services in medical complex envi-ronmentsrdquo in Proceedings of the 6th International Conferenceon Information Intelligence Systems and Applications (IISA rsquo15)Corfu Greece July 2015

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



concentration of users and intrinsic characteristics of net-work terminals and access pointsbase stations When con-sidering the implementation ofAAL environments and SmartHealth solutions we need to face scenarios that are complexin terms of wireless propagation due to the presence ofmultiple elements such as furniture building structure orpeople which can give rise to strong fading effects in anonuniform manner Moreover user density also affectsinterference levels with density values that can exhibit strongvariations as a function of scenario location or time ofanalysis particularly in the case of interaction with wearabledevices wireless body area networks or device-to-deviceconnections Accurate estimations of these interference levelswill help to avoid possible communication errors which insome cases like some e-Health applications where medicalsensors take part could be critical [7]

Wireless channel characterization in terms of usefulreceived signal levels for a given set of connections as wellas for potential interfering connections can be a challeng-ing task due to the large size of scenarios the existenceof multiple frequency dependent materials and inherentvariability of mobile connections given by the movementof potential scatterers In this sense several approachescan be followed from empirical estimations which pro-vide results at low computational cost with low precision(requiring site-specific calibration measurements) to full-wave simulation techniques which provide high accuracy atvery large computational cost As a midpoint deterministictechniques such as ray launching can provide adequateaccuracy while reducing computational cost As the scenariosize grows ray launching techniques increase their com-putational cost In this sense approaches such as sourcedefinition based on Huygens box approximation or the com-bination of 3D ray launching with neural networks reducethe computational effort when the size of the scenario grows[8 9]

In this paper novel approximation based on the com-bination of an in-house developed 3D ray launching codeand a collaborative filtering (CF) technique of bidimensionalcalculation points is used to analyze the performance ofwireless channels emulating AAL scenarios The studiedscenarios have an inherent complexity and a large numberof users and interferers The aim of the proposed method isto provide optimal deployment strategies formassive wirelesssystem and wireless sensor networks The application ofthe presented method in combination with new optimizingalgorithms [10ndash12] will enhance the WSNs performance inAAL and context-aware environments

2 Hybrid Simulation Technique

In this section the description of the in-house developed 3Dray launching code combined with a collaborative filteringapproach is presented This hybrid method will be used toanalyze the behavior of ZigBee wireless sensor networks(WSNs) in AAL environments with the aim of drasticallyreducing the computational time required to obtain accuratesimulation results

21 3D Ray Launching The 3D ray launching code is basedon Geometrical Optics (GO) and the Uniform Theory ofDiffraction (UTD)The principle of the algorithm is that raysare launched in a solid angle with input angular resolutionof parameters 120579 and Φ Electromagnetic phenomena such asreflection refraction and diffraction are taken into accountWe also consider the properties of all the obstacles within theenvironment considering the conductivity and permittivityof their different materials at the frequency of the systemunder analysis It is worth noticing that a grid is definedin the space and the ray launching parameters are storedat each cuboid during the propagation of each ray Theconfiguration parameters of the system are frequency powergain polarization and directivity of the transceivers bit rateangular resolution of the launching and diffracted rays andnumber of reflections The received power is calculated withthe sum of the electric field vectors inside each cuboid of thedefined mesh When a ray hits an obstacle new reflected andtransmitted rays are generated with new angles provided bySnellrsquos lawWhen a ray hits an edge a new family of diffractedrays is generated as we can see in Figure 1 with the diffractioncoefficients given by the UTD which are shown in (1) givenby [13 14]

119863perp=minus119890(minus1198951205874)

2119899radic2120587119896cot119892(

120587 + (Φ2 minus Φ1)

2119899)

sdot 119865 (119896119871119886+(Φ2 minus Φ1)) + cot119892(

120587 minus (Φ2 minus Φ1)

2119899)

sdot 119865 (119896119871119886minus(Φ2 minus Φ1)) + 119877

perp

0

sdot cot119892(120587 minus (Φ2 + Φ1)

2119899)119865 (119896119871119886

minus(Φ2 + Φ1))

+ 119877perp

119899 cot119892(120587 + (Φ2 + Φ1)

2119899)

sdot 119865 (119896119871119886+(Φ2 + Φ1))

(1)

where 119899120587 is the wedge angle 119865 119871 and 119886∓ are defined in [13]and 1198770119899 are the reflection coefficients for the appropriatepolarization for the 0 face or 119899 face respectivelyThe completeapproach has been explained in detail in [15] and it has beenused successfully in different complex indoor environments[16ndash22]

Two simulation configurations have been used to fit theCF method with results High Definition (HD) simulationsand Low Definition (LD) simulations HD simulations useparameters usually set to obtain accurate estimations inindoor environments On the contrary LD simulations pro-vide less accurate estimations but the computational cost ismuch lower The main parameters that have been used forboth LD and HD simulations are summarized in Table 1

22 Recommender Systems and Collaborative Filtering Rec-ommender Systems (RS) are a family of techniques used tomanage and understand the information created by users in


Diffraction

Refraction

Reflection

ReRefrac

















V2 V3

V1 V2 V3

V4


SV1 =

SV2 =

SV(Lminus2) =

minusLSV + 1


VL119881


VL119881





55

5 5

6 6 4

4

4 4

4

43

3

3 3

3

3

3

3

3

33

2

2

2

2

2

2

2

2

2

2

2

2

2

1

1

1

1

1

1

1

1

1

minus1

minus1

minus1

minus1



V1 V2 V4 V5

VPV1 V2 V4 V5



















Polycarbonate 3 02






7255m

905

m

2625

m(a)

Aisle

Bathroom

Kitchen

Box room

Study room

Living room

Bedroom 2

Bedroom 1

X

Y

ZRZED


(b)


(a) (b)






Transmitter

P14

P13

P12

P10 P9 P8 P7 P6P5

P4P11

P3

P2

P1

1 2 3 4 5 6 7 8 90

X-distance (m)

0

1

2

3

4

5

6

7

Y-d

istan

ce (m

)





Rece

ived

pow

er (d

Bm)








2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)




1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED




minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)




Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Diffraction

Refraction

Reflection

ReRefrac

















V2 V3

V1 V2 V3

V4


SV1 =

SV2 =

SV(Lminus2) =

minusLSV + 1


VL119881


VL119881





55

5 5

6 6 4

4

4 4

4

43

3

3 3

3

3

3

3

3

33

2

2

2

2

2

2

2

2

2

2

2

2

2

1

1

1

1

1

1

1

1

1

minus1

minus1

minus1

minus1



V1 V2 V4 V5

VPV1 V2 V4 V5



















Polycarbonate 3 02






7255m

905

m

2625

m(a)

Aisle

Bathroom

Kitchen

Box room

Study room

Living room

Bedroom 2

Bedroom 1

X

Y

ZRZED


(b)


(a) (b)






Transmitter

P14

P13

P12

P10 P9 P8 P7 P6P5

P4P11

P3

P2

P1

1 2 3 4 5 6 7 8 90

X-distance (m)

0

1

2

3

4

5

6

7

Y-d

istan

ce (m

)





Rece

ived

pow

er (d

Bm)








2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)




1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED




minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)




Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











V2 V3

V1 V2 V3

V4


SV1 =

SV2 =

SV(Lminus2) =

minusLSV + 1


VL119881


VL119881





55

5 5

6 6 4

4

4 4

4

43

3

3 3

3

3

3

3

3

33

2

2

2

2

2

2

2

2

2

2

2

2

2

1

1

1

1

1

1

1

1

1

minus1

minus1

minus1

minus1



V1 V2 V4 V5

VPV1 V2 V4 V5



















Polycarbonate 3 02






7255m

905

m

2625

m(a)

Aisle

Bathroom

Kitchen

Box room

Study room

Living room

Bedroom 2

Bedroom 1

X

Y

ZRZED


(b)


(a) (b)






Transmitter

P14

P13

P12

P10 P9 P8 P7 P6P5

P4P11

P3

P2

P1

1 2 3 4 5 6 7 8 90

X-distance (m)

0

1

2

3

4

5

6

7

Y-d

istan

ce (m

)





Rece

ived

pow

er (d

Bm)








2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)




1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED




minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)




Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation




















Polycarbonate 3 02






7255m

905

m

2625

m(a)

Aisle

Bathroom

Kitchen

Box room

Study room

Living room

Bedroom 2

Bedroom 1

X

Y

ZRZED


(b)


(a) (b)






Transmitter

P14

P13

P12

P10 P9 P8 P7 P6P5

P4P11

P3

P2

P1

1 2 3 4 5 6 7 8 90

X-distance (m)

0

1

2

3

4

5

6

7

Y-d

istan

ce (m

)





Rece

ived

pow

er (d

Bm)








2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)




1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED




minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)




Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










7255m

905

m

2625

m(a)

Aisle

Bathroom

Kitchen

Box room

Study room

Living room

Bedroom 2

Bedroom 1

X

Y

ZRZED


(b)


(a) (b)






Transmitter

P14

P13

P12

P10 P9 P8 P7 P6P5

P4P11

P3

P2

P1

1 2 3 4 5 6 7 8 90

X-distance (m)

0

1

2

3

4

5

6

7

Y-d

istan

ce (m

)





Rece

ived

pow

er (d

Bm)








2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)




1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED




minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)




Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Transmitter

P14

P13

P12

P10 P9 P8 P7 P6P5

P4P11

P3

P2

P1

1 2 3 4 5 6 7 8 90

X-distance (m)

0

1

2

3

4

5

6

7

Y-d

istan

ce (m

)





Rece

ived

pow

er (d

Bm)








2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)




1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED




minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)




Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(a)

2385

241

2435

Freq

uenc

y (G

Hz)

15 30 45 60 75 900

Time (s)



(b)




1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZED




minus90

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

Rece

ived

pow

er (d

Bm)

2 80 64Distance (m)




Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Sensitivity

10 20 30 40 50 60 700

Time (ns)

minus120

minus100

minus80

minus60

minus40Re

ceiv

ed p

ower

(dBm

)

(a)

Sensitivity

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

10 20 30 40 50 60 70 800

Time (ns)

(b)

Sensitivity

20 40 60 800

Time (ns)

minus120

minus100

minus80

minus60

minus40

Rece

ived

pow

er (d

Bm)

(c)







020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










020406080

100120140

24

68

12345Dela

y Sp

read

(ns)

X-dista

nce (m)

Y-distance (m)

0 20 40 60

80 100 120 140

46

8

45dista

nce (m




ZED

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)

ZR






119862 = BW log2 (1 +119878

119873) (2)





ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












ZED

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)

(a)



ZED

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)(b)


ZED ZED

ZED ZED

ZR

4 6 82X (m)

1

2

3

4

5

6

7

Y (m

)



(a)

ZED ZED

ZEDZED

ZR

1

2

3

4

5

6

7

Y (m

)

2 3 4 5 6 7 81X (m)



(b)





WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










WiFi

WiFi

ZR

1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

WiFi

WiFi

ZR

2 3 4 5 6 7 81X (m)

1

2

3

4

5

6

7

Y (m

)



(b)











MAE =sum119899

119894=1

1003816100381610038161003816119901119894 minus 1199031198941003816100381610038161003816

119899 (3)


1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










1

2

3

4

5

6

7Y

(m)

4 6 82X (m)



(a)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(b)

1

2

3

4

5

6

7

Y (m

)

4 6 82X (m)



(c)











minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














minus505

10152025

minus10 0 10 18

SNR

(dB)



256Kbps

32Kbps





5 Conclusions






Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Sim

ulat

ion

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 950Time in a 103 scale (s)


HDLDLD + CF

0102030405060708090

MA

E (d

B)

10 100 1000 10000 1000001Time in a log

10scale (s)







Acknowledgments


References


































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation

































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Research Article Performance Analysis of ZigBee Wireless …downloads.hindawi.com › journals ›...

Documents

Transcript of Research Article Performance Analysis of ZigBee Wireless …downloads.hindawi.com › journals ›...