Hybrid Radio-map for Noise Tolerant Wireless Indoor Localization
HHybrid Radio-map for Noise Tolerant WirelessIndoor Localization
Xiongfei Geng ∗ , Yongcai Wang † , Haoran Feng ‡ and Zhoufeng Chen ∗∗ China Waterborne Transport Research Institute, Beijing, P. R. China † Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, Beijing, P. R. China ‡ National Engineering Research Center of Software Engineering, Peking University, Beijing, P. R. China
Abstract —In wireless networks, radio-map based locating tech-niques are commonly used to cope the complex fading feature ofradio signal, in which a radio-map is built by calibrating receivedsignal strength (RSS) signatures at training locations in the offlinephase. However, in severe hostile environments, such as in shipcabins where severe shadowing, blocking and multi-path fadingeffects are posed by ubiquitous metallic architecture, even radio-map cannot capture the dynamics of RSS. In this paper, we intro-duced multiple feature radio-map location method for severelynoisy environments. We proposed to add low variance signatureinto radio map. Since the low variance signatures are generallyexpensive to obtain, we focus on the scenario when the lowvariance signatures are sparse. We studied efficient constructionof multi-feature radio-map in offline phase, and proposed feasibleregion narrowing down and particle based algorithm for onlinetracking. Simulation results show the remarkably performanceimprovement in terms of positioning accuracy and robustnessagainst RSS noises than the traditional radio-map method.
I. I
NTRODUCTION
For on-ship wireless sensor networks [1], [2], [3], one of themost important and desirable applications is to provide real-time location information for crews, passengers or facilities byusing the sensor network as a wireless locating infrastructure.When the ships seal on the sea, the location information willbe fundamental context for safety-oriented applications or on-ship facility management.For locating using sensor networks, wireless location tech-niques have attracted great research attentions in recent years.Various locating techniques have been developed using differ-ent implementation techniques, but it is still challenging to finda balance between the positioning accuracy and the systemcost. Some positioning techniques provide good positioningaccuracy, such as TOA (Time of Arrival) [5] or TDOA (TimeDifference of Arrival)[6] based localization methods, but thesemethods generally need special hardware, such as ultrasoundor acoustic transducers, which need additional hardware costs.Some other methods are inexpensive, such as RSS (RadioSignal Strength) based wireless location [7], because RSSinformation is free-of-charge. But these inexpensive methodsprovide only coarse-grained positioning accuracy.A way to increase the positioning accuracy by using RSS isto exploit a radio-map based locating method [8], [9], whichtrains the RSS fingerprints of all cared locations in an offlinecalibration phase, to construct a radio map ; then the onlinemeasured RSS of the target is searched in the radio-map to findthe location whose RSS signature matches best to the online RSS measurement. The location is chosen as the positionestimation of the target. Probabilistic radio-maps and Bayesianreasoning methods can be applied to improve the positioningaccuracy.In our previous works, we have studied both ultrasoundTOA-based locating systems [5], [10] and radio-map basedlocating systems in buildings [11]. But when we developedand tested these systems on ships, dramatic performancedegradation was found because of the signal blocking effectsby the metallic architecture of the cabins. The multi-path,shadowing, and blocking effects are serious. The receivedsignal strength can be very weak even when the receivers areclose to a transmitter but are not in line of sight (NLOS).To deal with this problem, in this paper, we propose efficientmethods to construct multi-feature radio-map to overcome thehostile environments for wireless localization.To be tolerant to the noise of RSS, we proposed hybridradio-map integrating both low-variance signature and the RSSsignature. More particularly, we deploy ultrasound beaconssparsely in the sensing field which contributes sparse, bylow-variance time-of-arrival (TOA) information of ultrasoundfrom the transmitter to receiver. We show that by integratingthese sparse low variance information into radio-map, it candramatically improve the positioning accuracy of Radio-mapbased positioning systems.Utilizing the lower variance signature, we proposed a newefficient method to offline calibrate hybrid radio-maps withoutthe pain of manual calibration. Then in the online phase,instead of simple matching algorithm, we propose to use thelow variance signature to narrow down the feasible spacefirstly, and then use particle filter based algorithm to efficientlyand accurately tracking the mobile targets. Since the lowvariance beacons are very sparse, only a little additional costsare needed, but the hybrid radio-map system can provide dra-matical improvement in positioning accuracy and reliability.The remainder sections are organized as following. Back-ground and related works are introduced in Section 2. Weintroduce the hybrid radio-map model construction and onlinetracking algorithm in Section 3. Simulation based evaluationresults are introduced in Section 4. Conclusions are drawn inSection 5. a r X i v : . [ c s . N I] D ec I. B
ACKGROUND AND R ELATED W ORKS
Using wireless networks as indoor locating infrastructure,there are different ways to utilize the wireless signal. Oneway is to use the propagation model of RF signal as a rangingreference. Some research works studied the RF attenuationmodel in indoor environments [7], so that distances from atarget to a set of beacons can be inferred from the amount ofRF attenuations. Then least square estimation or multilatera-tion methods [4] are applied to the distance set to estimatethe position of the target. Although this method is simpleto calculate, the positioning accuracy is coarse, because evenempirical propagation model cannot capture the dynamics ofindoor environments.
A. Radio-map Locating Method
To improve the positioning accuracy, pattern-matching based approach was proposed to model the diverse fadingsignatures of radio signal [8][12][13]. This method containsan offline and an online phase. In offline phase, n training lo-cations are selected in the sensing field, which are denoted by L = { l , l , · · · , l n } . Suppose there are m beacons (WiFi APsor wireless sensors) in the sensing field, which are denoted by B = { b , b , · · · , b m } . In the training phase, the RSS valuesof all beacons at each training location l i will be measuredover a period of time, so that a signal signature vector oflocation l i is constructed as r i = { r i, , r i, , · · · , r i,m } . Whenonly mean value of RSS is considered, r i,j represents theaverage RSS value from b j , j = 1 , · · · , m . When signaturedistribution is considered, r i,j can be probabilistic densityfunction (pdf) of RSS from b j . The signature vectors of alltraining locations are stored as a database, called radio-map ,denoted by R = { r , r , · · · , r n } .In the online positioning phase, a mobile target measuresits current RSS vector s = { s , s , · · · , s m } and finds thebest match (Euclidean distance in signal space) of s in R to estimate the position of the target. In mean value typeradio-map, matching can be conducted by Nearest Neighboralgorithm[13]. In pdf type radio-maps, maximum likelihoodestimation and Bayesian estimation can be applied. Whenradio-map is trained in fine granularity and the environmentsare not highly dynamic, the positioning accuracy of radio-mapbased method can be in 2-3 meters resolution.But the positioning accuracy may become worse in hostileenvironment such as in ship cabins, where the shadowing andmulti-path fading effects are severe and the RSS signatureschange over time. Another problem is that the radio-mapcalibration process is general time consuming and laborious,which generally needs deliberate training method [9]. B. Locating by Time of Arrival (TOA)
A more accurate approach is to utilize the speed differenceof signal propagation to measure distances from transmit-ters to receivers, so as to conduct indoor locating moreaccurately[5][6]. Ultrasound and acoustic signals are the gen-erally exploited low speed signals. In the case of measuringtime of arrival, the transmitter broadcasts low-speed signal (ultrasound or acoustic) and RF signal simultaneously. Thereceiver receives the RF signal to synchronize timer with thetransmitter and then measures the traveling time of the low-speed signal to estimate distance from the transmitter. Onlywhen a receiver j is within the communication range of thelow speed signal (denoted by R , and generally small) of thetransmitter i , can a distance d i,j be measured. When a set ofdistances, which is denoted by D i = { d i,j } are obtained, leastsquare estimation or multilateration is applied for position cal-culation. TOA-based positioning can provide centimeter levelpositioning accuracy[10]. But because the short transmissionrange of the low speed signals (ultrasound and acoustic), andthe requirement of more than three non-collinear distancesfor location estimation, TOA-based positioning requires densedeployment of TOA beacons, which poses high cost to thepositioning system.III. H YBRID R ADIO -M AP L OCATING M ETHOD
Note that the radio-map based and TOA-based wirelesslocating methods both have advantages and shortcomings. Wepropose a method to integrate their advantages and to avoidtheir shortcomings. Our proposed method is not specificallydesigned for integrating RSS and TOA signatures, it is actuallydesigned for integrating RSS with a low variance signaturesuch as TOA, time difference of arrival (TDOA) etc. There-fore, in the following model, we call the second signature lowvariance signature (LVS).Let’s consider a hybrid positioning system containing aset of RF beacons and some sparsely deployed LVS bea-cons in the sensing field. The RF beacons are denoted by B = { b , b , · · · , b m } and the LVS beacons are denoted by V = { v , v , · · · , v g } .In offline phase, some training locations L = { l , l , · · · , l n } are selected in the sensing field. At eachtraining location, after a training target listens to beaconsignals for a period of time. It can learn a set of beaconsignatures s l = { d ,l , · · · , d g,l , s ,l , · · · , s m,l } , where d i,j and s i,j are the LVS signature from v j and RSS signature from b j respectively. We consider the signatures are calculated bytaking average on the collected signatures in the training time.We assume the LVS signature is distance-based signature,such as TOA or TDOA. An identical variance δ is assumedfor all the LVS signatures. We store the multi-feature vectoras the signature of location l .In online phase, a target can online detect a set of beaconsignals s (cid:48) = { d (cid:48) , · · · , d (cid:48) g , s (cid:48) l , · · · , s (cid:48) m } . Note that, for thelimited communication range of beacons, many entries of s (cid:48) are zero. We design efficient algorithm to match s (cid:48) against theradio map to find a location l (cid:48) whose radio signature has theleast distance to s (cid:48) as the position estimation of the target.Since the mobile target has limited moving speed, its histor-ical track implies important clues for its future position. There-fore, based on the online hybrid radio-map locating scheme,we designed particle filter algorithm to more accurately trackthe movements of the mobile targets. F beacon
LVS beacon
Detec/on range of LVS signal
Fig. 1. An example of hybrid indoor locating system
Track-‐based Beacon Signature Associa5on
Fast Hybrid Radio-‐map Calibra5on
Priori5zed posi5on candidates lis5ng
Par5cle filter based target tracking
Hybrid radio-‐map
HIstorical Posi5ons of targets
Online signature measurement
Offline signature collec5on output
Offline phase
Online phase
Fig. 2. Routine of proposed methods for hybrid indoor locating system
We use example in Fig.1 to illustrate the scenario of hybridindoor locating system. The dashed curve is the communica-tion range of the LVS beacons. Note that the communicationranges of beacons are irregular in indoor environments.The overview of this paper for calibrating, locating andtracking using hybrid radio-map is shown in Fig.2. With theadvantage of hybrid beacons, we at first present a method tofast calibrate the hybrid radio-map. It is based on the controlledtracks of the target movements. Then in the online phase,we presented an algorithm to assign radio signature differentpriorities to list a set of possible positions of the targets bymatching the online measured signatures of targets againstthe radio map. We then present particle filter based targettracking algorithm to utilize the historical position informationto accurately track the mobile targets.
A. Fast Calibration of Hybrid Radio-map
Radio-map calibration is known time-consuming and labo-rious, which is a major limitation of radio-map based locating.By utilizing the advantages of hybrid beacons, we present anefficient method for fast radio-map calibration.The LVS beacon, whether using TOA or TDOA signaturecan provide ranging information from the target to the beaconwith good precision. So that in the offline calibration phase,instead of manually assigning location labels to the signalsignatures, we design controlled tracks for the training targetsto collect hybrid signatures and then use algorithm to calculatethe location labels for the hybrid signatures. The working flowof fast calibration method is as following:1)
Feature points selection: we select some feature points in the sensing field, which are denoted F = { f , f , · · · , f h } . Let h denote the number of featurepoints. The rule to select the feature points is thatthe path between two neighboring feature points is adirected line. We manually assign location labels to thehybrid signatures of these feature points.2) Controlled training paths:
A training target is movedalong the paths that connecting the feature points. Eachsegment of its movement is a line, and the whole pathcan be represented by a set of feature points p i = { f i, , f i, , · · · , f i,e i } , where f i,j ∈ F , j = 1 , · · · , e i .Since we know the start point and end point of eachline segment, we can infer the location labels of all theintermediate points.3) LVS-aided fast calibration: to calibrate the locationlabel of each point on the line segment, we adopttwo methods based on whether the point is in thecommunication range of the LVS beacon.
Case 1 : When the point whose location label need to becalibrated is in the communication range of a LVS beacon,since we know the starting point and ending point of thisline segment and know the distance from this point to a LVSbeacon (provided by the LVS signature), we can calculate thelocation label of this point precisely. Let ( x s , y s ) and ( x e , y e ) be the coordinates of the starting point and end point of theline segment that the point is on. The problem of calibratingthe location label of the point is to find a point on this linewho has distance d i to the LVS beacon v i . This point can becalculated by: (cid:40) y = y e − y s x e − x s x + y s x e − y e x s x e − x s (cid:113) ( y − y v i ) + ( x − x v i ) = d i (1) LVS beacon f f f p p d i, v i Feature point
Controlled training path
Communica7on range of LVS beacon
Fig. 3. Example of LVS-aided fast calibration method.
This equation array generally contains two ambiguous so-lutions. But from the moving direction of the target, we caneasily disambiguate to determine an unique solution. As shownby the example in Fig.3, p is a point on a line segment thatis in the communication range of a LVS beacon. We want toassign location label to the RSS signatures measured at thispoint. There are two intersecting points of the line with thecircle centered at v i with radius d . They are both solutions ofEqn.(1). But since the target is moving from f to f , the firstpoint with distance equal to d is the location of p , whichdisambiguate the problem. ase 2 : When the point whose location label need tobe calibrated is out the communication range all LVS bea-cons, because we know the starting point and ending pointof the line segment, we can infer its location reasonablyby linear interpolation. Suppose the training target movesin constant speed. If the times at the starting point, end-ing point and the point to be calibrated are t s , t e , t p re-spectively, then the location of the point can be estimatedby (cid:16) x s + ( t p − t s )( x e − x s ) t e − t s , y s + ( t p − t s )( x e − x s ) t e − t s (cid:17) . p in Fig.3 issuch a case. Its position can be inferred by interpolation basedon the location of f and f . B. Prioritized Online Positioning Algorithm
After the hybrid radio-map is calibrated in the offline phase,the system turns to online phase to track the positions ofmobile targets. A target can detect a set of beacon signals s (cid:48) = { d (cid:48) , · · · , d (cid:48) g , s (cid:48) , · · · , s (cid:48) m } . Because of the sparse deploy-ment of the LVS beacons and there limited communicationrange, most of the LVS entries are zero. In case the target isout the communication range of all the LVS beacons, all theLVS entries are zero.Because the LVS signature has much less variance than thatof the RSS signature, only if there is one LVS signature in s (cid:48) , the LVS signature will provide very valuable informationfor target position estimation. Instead of simply matching theonline measured signature in the radio-map, we present aprioritized approach to always process the LVS signatures first.1) Characterize feasible region by LVS signature.
A non-zero LVS signature d (cid:48) i indicates that the distance fromthe target to v i is a random variable with distribution N ( d (cid:48) i , δ ) . Since δ is small, we can think the target is ona circle with distance at most d (cid:48) i + 3 δ , at least d (cid:48) i − δ around the beacon v i . This region is called feasibleregion , where the target must locate in. The feasibleregion can dramatically narrow down the searchingspace for target position.2) Find possible locations of target in the feasible region.
In the second step, the online measured RSS signatureis compared to the trained RSS signature of locationsin the feasible region. The locations whose trained RSSsignatures match well with the online measurement iselected as possible positions of the target. Let L e denotethe possible positions of the target. Let set F include thelocations in the feasible region. For all l ∈ F , l ’s trainedRSS signature is compared to the RSS signature in s (cid:48) .If (cid:107) s i,l − s (cid:48) i (cid:107) < H , i.e., the RSS distance is less thana threshold, l is added into the possible position set L e . L e = ∪ l, ∀ l ∈ F , if (cid:32) m (cid:88) i =1 (cid:107) s i,l − s (cid:48) i (cid:107) < H (cid:33) (2) H is a RSS-distance threshold which rules out the loca-tions whose RSS signatures don’t match RSS signatureof s (cid:48) .An example of the prioritized positioning process is shownin Fig.4. The green circle is the feasible region characterized by the LVS signature. The color in the map shows whetherthe RSS-distance of the location is smaller than H . Only thelocations in brown color have RSS-distance smaller than H .Among them, only the locations which are also in the feasibleregion are added into the possible target position set, whichare marked by a “ (cid:88) ”. d i Feasible region characterized by LVS signature ✔ ✔ ✔ ✔ RSS-‐distance larger than H RSS-‐distance smaller than H Possible loca;ons of target
Fig. 4. Example of prioritized online positioning algorithm.
C. Particle Filter based Target Tracking
Note that the online positioning algorithm provide a set ofpossible positions instead of a unique position estimation. It isdesigned in this way because the RSS signature is unreliable. Ifwe determine a unique location using the RSS signature, largelocating error maybe incurred. Therefore, we keep a set ofpossible positions at each time and proposed particle filteringbased algorithm to find the optimal moving track of the target.Let’s denote the possible locations of the target calculated attime t is in location set L ( t ) .
1) Particle Filter:
Since every target has an ID, we onlyneed to consider the case of tracking one target. At any time t , we generate K particles (or candidate trajectories), in thepossible locations of the target. Let’s denote the k th particle z k [ t ] . At the next time instant t + 1 , we generate m > K position candidates uniformly at random in L ( t ) for z k [ t +1] . We now have mK candidate trajectories (particles). Pickthe K particles with the best cost functions to get the set z k [ t + 1] , k = 1 , , K , where the cost function is to specifiedshortly. Repeat until the end of the time interval of interest.The final output is simply the particle (trajectory) with the bestcost function.
2) Cost Function:
The cost function is designed based onthe fact that the mobile target has restriction in its movingspeed (a target will not change speed suddenly). Therefore,we proposed an cost function that penalizes changes in thevector velocity. When a candidate position z k ( t + 1) is chosenfrom the current possible position set L ( t + 1) , the incrementin position z k ( t + 1) − z k ( t ) is an instantaneous estimate ofthe velocity vector at time t . The cost at time t is thereforedefined as the norm squared of the difference between theelocity vector estimates at time t and t − . This is: c k [ t ] = (cid:107) ( z k [ t + 1] − z k [ t ]) − ( z k [ t ] − z k [ t − (cid:107) = (cid:107) z k [ t + 1] + z k [ t − − z k [ t ] (cid:107) (3)IV. S IMULATION AND N UMERICAL R ESULTS
We conducted extensive simulations to verify the advantagesof using hybrid radio-map and particle filter than than thetraditional RSS-based radio-map method.
A. Simulation Settings
The simulation is conducted in Matlab 2012. We providethe code online at [14]. We simulate the an environment of120m*80m, in which 10 RSS beacons are randomly deployed.The radio propagation model used in simulation is [15]: P r ( d ) = P t − P l ( d ) − η log (cid:18) dd (cid:19) + N (0 , σ ) (4)We choose P t = 100dbm, d = 1 m , η = 3 , and σ = 3 inthe following simulation results. Note that since σ = 3 , thevariance of RSS can be more serious than general indoor RSSmodels [13].Six TOA beacons are deployed in grid topology. They areactivated only when TOA signature is used. The communica-tion radius of each TOA beacon is set to 25m, so that eachpoint in the sensing field can almost be covered by one TOAbeacon. In the offline phase, the training points are selectedin m ∗ m granularity. Hybrid or RSS radio-maps are trainedrespectively based on the positioning algorithms. In onlinephase, a target moves in the sensing field following a sin-wavepath. (cid:26) x = WT ty = H (cid:0) sin (cid:0) πxW (cid:1) + 1 (cid:1) (5) W and H are the width and height of the sensing field (equalto 120 and 80 respectively in our setting). T is the lengthof the simulation , so that the target can finish a sin-wave inperiod T .We evaluated and compared three kinds of positioningalgorithms:1) Hybrid radio-map with particle filter , in which TOAbeacons are activated. Fine-grained hybrid radio-mapare trained offline (in 1m*1m granularity) and locationalgorithms introduced in Section III-B, Section III-C areevaluated.2)
Hybrid radio-map without particle filter.
TOA beaconsare activated. RSS radio-map are trained offline in1m*1m granularity. Location algorithm introduced insection III-B is used by limiting the number of posi-tioning candidates to 1, without using particle filter.3)
RSS radio-map without particle filter.
TOA beacons areinactivated. RSS radio-map are trained offline in 1m*1mgranularity. The position with the least RSS signaturedifference to the online measured RSS is estimated asposition of the target.
B. Effectiveness of Hybrid Radio-map and Particle Filter
We at first visually show the effectiveness of using hybridradio signature.
1) Narrow down feasible region:
One important contribu-tion of the TOA signature is to provide low variance positionestimation and dramatical search space narrowing down. Asshown in Fig.6a, the red square points are possible positioncandidates generated by least RSS-distance. We can see theestimated possible positions are highly scattered in the sensingfield for the unreliability of the RSS signal. Fig.6b shows howthe TOA-signature helps to narrow down the feasible region.Since the possible positions must be in the feasible region (thecircle area), the possible positions are filtered.
PositionCandidatesbypureRSSsignaturematching
Feasible regioncharachterized byTOA signaturesPosition candidates narroweddown by feasible region
Fig. 6. Search space narrow down by TOA signature
2) Improve the positioning accuracy:
Fig.5 compares dif-ferent algorithms to show the improvement of positioningaccuracy by the hybrid radio-map and particle filter algorithms.The red sin-wave curves in the figures shows the ground truthof the target movement. Fig.5(a) illustrates the target trackingperformance when using hybrid radio-map plus particle filtertracking algorithm. The algorithm can provide accuracy targettracking performance. Fig.5(b) shows the tracking resultswhen only hybrid radio-map is used without the particlefilter algorithm. Instead of particle filter, at each step, thecandidate position which matches best to the hybrid signatureof measurement is chosen as the position estimation. We cansee the positioning performance degrades much than that inFig.5(a). Fig.5c shows the tracking results when only RSS-based radio-map is used without using TOA signature norparticle filter. At each step, the candidate position in theradio-map which matches best to RSS-signature of targetis chosen as the position estimation. We can see that thetracking performances become much worse than the prior twoapproaches.Fig.7 uses cumulative probability distribution of position-ing error to further illustrate the locating performance im-provement achieved by hybrid radio-map and particle filteralgorithms. The results are based on the average positioningerrors of 20 simulations. It show that the method of hybridradio-map plus particle filter performs the best, which is a littlebetter than the method of using hybrid radio-map but withoutparticle filter. The difference of them is that the latter methodmay has a small portion of results having large positioningerror, which is not robust. Both of these two methods aremuch better than the traditional methods using RSS-based
20 40 60 80 100 12001020304050607080 ground truthestimated tranjactory (a) Hybrid Radio-map with Particle Filter ground truthestimated tranjactory (b) Hybrid Radio-map without ParticleFilter ground truthestimated tranjactory (c ) RSS-based Radio-map without usingTOA signature and Particle Filter
Fig. 5. Comparing tracking performances of different algorithms
Positioning Error (m) C u m u l a t i ve p r ob a b ili t y d i s t r i bu t i on Empirical CDF of positioning error
Hybrid Radio−map with Particle FilterHybrid Radio−map without Particle FilterRss−based Radio−map without Particle Filter
Fig. 7. Cumulative probability distribution of positioning errors for differentpositioning algorithms. radio-map. The results show the significance of the sparselow-variance signature for the improvement of positioningaccuracy. V. C
ONCLUSION
This paper presents hybrid radio-map method for improvethe positioning accuracy of RSS-based wireless indoor lo-calization. It presents efficient methods to utilize the sparse,low variance signature to construct hybrid radio-map, andpresents particle filter based algorithm for accurate onlinetarget tracking. Simulation results verified that by using verylimited TOA beacons, the hybrid radio-map method candramatically improve the positioning accuracy of wirelesslocation systems. We will conduct hardware experiments andsystem development in our future work.A
CKNOWLEDGMENT
This work was supported in part by National NaturalScience Foundation of China Grant 61202360, Major Scienceand Technology Project on Transportation Informatics forCommunication Networks of Internet of Ships (2012-364-222-203), and the National Basic Research Program of China Grant2011CBA00300, 2011C-BA00302. R
EFERENCES[1] B.-G. Paik, S.-R. Cho, B.-J. Park, D. Lee, B.-D. Bae, andJ.-H. Yun, “Characteristics of wireless sensor network for full-scale ship application,”
Journal of Marine Science and Technology ,vol. 14, no. 1, pp. 115–126, Mar. 2009. [Online]. Available:http://link.springer.com/article/10.1007/s00773-008-0038-x[2] H. Kdouh, H. Farhat, T. Tenoux, C. Brousseau, G. Zaharia, G. Grun-felder, Y. Lostanlen, and G. El Zein, “Double directional channelcharacterization on board ships,” Jun. 2013, pp. 224–229.[3] L. Mu, “A hybrid network for maritime on-board communications,”in , 2012, pp. 761–768.[4] G. Han, H. Xu, T. Q. Duong, J. Jiang, and T. Hara, “Localizationalgorithms of wireless sensor networks: a survey,”
TelecommunicationSystems , vol. 52, no. 4, pp. 2419–2436, Apr. 2013. [Online]. Available:http://link.springer.com/article/10.1007/s11235-011-9564-7[5] J. Zhao and Y. Wang, “Autonomous ultrasonic indoor tracking system,”in
International Symposium on Parallel and Distributed Processing withApplications, 2008. ISPA ’08 , Dec. 2008, pp. 532 –539.[6] F. Gustafsson and F. Gunnarsson, “Positioning using time-differenceof arrival measurements,” in , vol. 6, 2003, pp. VI–553–6 vol.6.[7] K. El-Kafrawy, M. Youssef, A. El-Keyi, and A. Naguib, “Propagationmodeling for accurate indoor WLAN RSS-Based localization,” in
Ve-hicular Technology Conference Fall (VTC 2010-Fall), 2010 IEEE 72nd ,2010, pp. 1–5.[8] I. Haque and C. Assi, “Profiling-based indoor localization schemes,”
IEEE Systems Journal , vol. Early Access Online, 2013.[9] P. Scholl, S. Kohlbrecher, V. Sachidananda, and K. Van Laerhoven, “Fastindoor radio-map building for RSSI-based localization systems,” in ,2012, pp. 1–2.[10] Y. Wang, J. Zhao, and T. Fukushima,
LOCK: A Highly Accurate, Easy-to-Use Location-Based Access Control System , 2009.[11] V. Zheng, J. Zhao, Y. Wang, and Q. Yang, “HIPS: a calibration-lesshybrid indoor positioning system using heterogeneous sensors,” in
IEEEInternational Conference on Pervasive Computing and Communications,2009. PerCom 2009 , Mar. 2009, pp. 1 –6.[12] L. Ni, Y. Liu, Y. C. Lau, and A. Patil, “LANDMARC: indoor locationsensing using active RFID,” in
Proceedings of the First IEEE Interna-tional Conference on Pervasive Computing and Communications, 2003.(PerCom 2003) , 2003, pp. 407–415.[13] P. Bahl and V. Padmanabhan, “RADAR: an in-building RF-based userlocation and tracking system,” in