International Journal of

Geo-Information

Article
Integrated WiFi/PDR/Smartphone Using an Adaptive
System Noise Extended Kalman Filter Algorithm for
Indoor Localization
Xin Li 1,2, *, Jian Wang 1 , Chunyan Liu 1 , Liwen Zhang 1 and Zhengkui Li 2
1 Jiangsu Key Laboratory of Resources and Environment Information Engineering, Xuzhou 221116, China;
[email protected] (J.W.); [email protected] (C.L.); [email protected] (L.Z.)
2 School of Computer Science and Technology, China University of Mining and Technology, Xuzhou 221116,
China; [email protected]
* Correspondence: [email protected]; Tel.: +86-152-5203-9940

Academic Editors: Georg Gartner, Haosheng Huang and Wolfgang Kainz

Received: 19 October 2015; Accepted: 22 January 2016; Published: 1 February 2016

Abstract: Wireless signal strength is susceptible to the phenomena of interference, jumping, and
instability, which often appear in the positioning results based on Wi-Fi field strength fingerprint
database technology for indoor positioning. Therefore, a Wi-Fi and PDR (pedestrian dead reckoning)
real-time fusion scheme is proposed in this paper to perform fusing calculation by adaptively
determining the dynamic noise of a filtering system according to pedestrian movement (straight
or turning), which can effectively restrain the jumping or accumulation phenomena of wireless
positioning and the PDR error accumulation problem. Wi-Fi fingerprint matching typically requires a
quite high computational burden: To reduce the computational complexity of this step, the affinity
propagation clustering algorithm is adopted to cluster the fingerprint database and integrate the
information of the position domain and signal domain of respective points. An experiment performed
in a fourth-floor corridor at the School of Environment and Spatial Informatics, China University of
Mining and Technology, shows that the traverse points of the clustered positioning system decrease
by 65%–80%, which greatly improves the time efficiency. In terms of positioning accuracy, the average
error is 4.09 m through the Wi-Fi positioning method. However, the positioning error can be reduced
to 2.32 m after integration of the PDR algorithm with the adaptive noise extended Kalman filter (EKF).

Keywords: indoor positioning; affinity propagation clustering; feature fusion; pedestrian dead
reckoning; multi-sensor fusion

1. Introduction
Indoor navigation has become an essential technology in a number of applications, such as in a
supermarket as a shopping guide, a fire emergency service for navigation, or a hospital patient for
tracking. However, some successful technologies that are similar to the Global Navigation Satellite
System (GNSS) [1–3] are unsuitable for indoor navigation. Real-time indoor positioning via existing
technologies remains a challenge, and this is a bottleneck in the development of indoor location-based
services (LBSs) [4].
The solution for indoor positioning is increasingly regarded as being based on multiple integrated
technologies—e.g., Wi-Fi, ZigBee, inertial navigation systems (INSs), and laser scanning systems (LSSs).
Each has its shortcomings, but an integrated system can combine the advantages of several of these
technologies. Pahlavan and Li reviewed the technical aspects of existing technologies for wireless
indoor location systems [5]. There are two main hardware layouts that can be used in an indoor
situation: (1) a sensor network, such as a Wi-Fi or ZigBee system [6–8]; and (2) self-contained sensors,

ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5020008 www.mdpi.com/journal/ijgi

ISPRS Int. J. Geo-Inf. 2016, 5, 8 2 of 19

such as gyroscopes, accelerometers, or magnetometers [9–12]. However, the stringent demands of

reliable and continuous navigation in indoor environments are unlikely to be achievable using a single
type of layout, and developing a hybrid scheme for reliable and continuous positioning is therefore a
core prerequisite for real-time indoor navigation [13–15].
Developing a hybrid scheme for real-time indoor navigation is a more effective practice [16].
Li et al. presented a Wi-Fi-aided magnetic matching (MM) algorithm for indoor pedestrian navigation
with consumer portable devices. This algorithm reduces both the mismatching rate and computational
load of MM by using Wi-Fi positioning solutions to limit the MM search space [17]. Masiero et al.
proposed a particle filtering method based on the integration of information provided by the inertial
navigation system measurements, the radio signal strength of a standard wireless network, and
the geometrical information of the building [18]. Aicardi et al. integrated the data captured from
mobile phone cameras into indoor pedestrian dead reckoning and used image matching to achieve
positioning [19]. Saeedi et al. proposed an approach based on context-aware personal navigation,
the basic idea being that mobile navigation services can provide different services based on different
contexts—where contexts are related to the user’s activity and the device placement [20].
However, most existing methods may need additional information such as images and magnetic
fields, which can not only increase the volume and power consumption of the system but also be
more easily influenced by the external environment. Moreover, most existing methods require large
data calculation, which is suitable for post-processing analysis, and high operational capability of the
processor, which is unsuitable for an application with a low-cost processor. This paper proposes a
novel data fusion framework by using an adaptive extended Kalman filter (EKF) to integrate Wi-Fi
localization with PDR.
The fingerprint-based Wi-Fi indoor positioning method includes offline sampling and real-time
matching and positioning. Offline sampling selects a number of RPs (reference points) in the target zone
to obtain the fingerprint database in the target environment zone by sampling Wi-Fi signal information,
including the location information and signal strength of RPs, the MAC address information of APs
(Access Points), and other data; in the real-time matching stage, when receiving real-time Wi-Fi signal
data, the user target device will conduct a matching calculation between the data and the fingerprint
database and then estimate the current position of the target user with the location of a known
reference point. The fingerprint positioning methods usually include the KNN (K-Nearest-Neighbor)
method [21], HORUS [22–24], etc.
The basic idea of the KNN algorithm is to classify the target into the nearest sample class in the
feature space. Owing to the complex and varied features of the indoor environment, the complexity of
the environment cannot be adequately characterized by using the mean signal strength as the basic
element of the fingerprint database. Therefore, Youssef proposed the HORUS positioning system to
conduct Gaussian modeling on the signal intensity distribution on the RPs and establish a fingerprint
database based on the average value and variance of the signal strength. However, if the points to be
calculated are too many, such as more than 1000, whether based on KNN or probability distribution, the
time consumption will increase. The algorithm based on probability distribution will especially take
more time, thus seriously lowering the timeliness of positioning; therefore, to improve the timeliness
and accuracy of the system, the data to be matched must be reduced.
The clustering process of a positioning fingerprint database may effectively reduce the data size of
the respective points to be searched by the system. The clustering feature selection can be based on the
strongest AP [25] or the distance between the RP signals [26]. The clustering process using the strongest
AP as the feature cannot significantly reduce the positioning calculation because of the shadow fading
effect caused by the strong reflection of the wireless signal in indoor environments [27]; the one using
the distance between signals as the feature can effectively avoid this situation [28]. However, the
ultimate goal of a positioning algorithm is to determine the location of an unknown point, so the use of
the distance between the signals cannot well reflect the relationship between the positions, which may
result in low accuracy of positioning. This paper integrates the signal feature and position feature of
ISPRS Int. J. Geo-Inf. 2016, 5, 8 3 of 19

the respective points in fingerprint database as the clustering feature and introduces the APC (Affinity
Propagation Clustering) method for clustering. Compared with the results of K-means, the class
division by the APC method is closer to the spatial structure division, and the class distribution is more
uniform. To further improve the stability and reliability of indoor positioning, a Wi-Fi signal and PDR
real-time fusion scheme is put forward. The advantage of a positioning method based on Wi-Fi signal
strength is that it can obtain the absolute coordinate figure. However, because the wireless signal is
easily blocked by obstacles and disturbed by the multipath effect and other factors, positioning failure
or instability phenomena such as location jumping or clustering may occur during the positioning; the
dead-reckoning system based on the inertial measurement unit can achieve relatively higher precision
position calculation, but the positioning error will rapidly increase with increasing walking distance.
Accordingly, a positioning model of dead reckoning and Wi-Fi fusion is established. This scheme
can adaptively determine the filtering system dynamic noise to conduct the EKF fusion calculation
according to the pedestrian movement (straight or turning), which can improve the stability of the
wireless positioning while restraining the error accumulation caused by dead reckoning.

2. The Wireless Positioning Technology Based on Field Strength

2.1. The Feature Extraction Integrating the Distance and Signal Information
Collection B = {b1 , b2 , ..., bn } comprises all APs within the fingerprint positioning area, where
all artificially planned RPs also compose Collection L = {l1 , l2 , ..., lm }. The element li (1 ď i ď m) in
Collection L consists of two pieces of data, one of which is represented by Gi = (xi , yi ), which is the
geographic coordinate of this point, and the other indicated by Vi = (vi.1 , vi.2 , ..., vi.n ), which is the
signal strength vector of each AP received at the point, where vi.j (1 ď I ď m, 1 ď jď n) represents the
intensity of the signal received in the spatial position associated with li from b j .
In the extraction of clustering feature, the distance in the position domain and signal domain
of the sampling points must be used. The signal distance is similar to the spatial distance; thus, the
Euclidean distance between the two groups of signals is defined as the signal distance in this paper,
as follows: c´
řn ¯2
vi,k ´ v j,k
˘ ||vi , v j ||
“ k “1
`
SigDis li , l j “ (1)
m_ij m_ij
where m_ij refers to the number of APs with nonzero signal strength received at both locations related
to li and lj . Typically, the value of m_ij is much larger when the positions related to li and lj are near
each other, and vice versa. Because the value of m_ij can reflect the spatial distance relationship between
the respective points in some degree, the introduction of variable m_ij can increase the signal distance
value between long-distance points and reduce the signal distance value between the short-distance
points when calculating the signal distance. The geographical distance between two sampling points
is defined as
` ˘
GeoDis li , l j “ ||Gi , Gj || (2)

Because the dimensions of the signal domain distance and position domain distance are
different, the distance used in the test is the Min-Max normalized result to integrate the features.
Although the trends of both kinds of distance in numerical values are almost the same—that is,
the numerical distribution of the signal distance and the spatial distance tends to be the same after
normalization—there is more noise in the signal domain distance owing to the instability of the signal.
To remove the effect of noise, the fusion feature uses the position domain distance to restrain the signal
domain distance, which is defined as:

MixDispli , l j q “ NorpSigDisnew pli , l j qq ˆ NorpGeoDispli , l j qq (3)

ISPRS Int. J. Geo-Inf. 2016, 5, 8 4 of 19

where Nor is the Min-Max normalized function. The fusion feature obtained by modifying the signal
distance with geographic position can better reflect the geographic relationship between sampling
points, and the clustered class distribution is more uniform and closer to spatial structure division.

2.2. Affinity Propagation Clustering

In the usual clustering algorithms, cluster centers are iteratively chosen and optimized to minimize
the quadratic sum of the distance between other members of the class and the cluster center. The
classic K-means algorithm, for example, must provide a series of initial cluster centers when starting.
Moreover, the algorithm is quite dependent on the initial value of the system and easily falls into
local extrema. Instead, in the APC [29], all points are connected in a network, in which each node can
be potentially considered as a cluster center. Through iteration, attraction and attribution messages
are continuously transmitted and received between the points, thus expanding the gaps between the
center point and the attachment points and finally determining the center point. Compared with the
K-means method, APC has a faster convergence rate and a lower absolute average error and average
variance [29].
The APC input information is the similarity matrix Sm˚m between m points, and the fused feature
MixDis is used in this paper. The similarity matrix Sm˚m represents the characteristic matrix for the
signal distance and the spatial distance between m APs with the calculation of every value made
according to the fusion feature, MixDis(i, j) in Formula 3. Moreover, s(i, j) represents the fusion feature
of the ith AP and the jth AP, also representing the fused value of the spatial distance and the signal
distance between two APs. Generally, a median in Sm˚m is chosen as a reasonable value to represent
s(k, k); namely, the median in line k of S is selected as the initial value of s(k, k). The calculation of the
message transmission between two reference points then follows, which is the core of the algorithm:
attraction message r(i, j) and attribution message a(i, j). Here, the attraction message r(i, j) is passed
from point i to point j to indicate the reliability of point j as the cluster center of point i; the attribution
message a (i, j) is passed from point j to point i to indicate the reliability of point i as the cluster center
of point j.

(1) Attraction message r(i, j)

The transmission from respective point li to respective point lj shows the attraction accumulation
of lj to li as the cluster center under the role of respective points except for li , and the formula is
as follows:
(
rpi, jq “ spi, jq ´ max api, j1 q ` spi, j1 q j1 ‰ j (4)

where s(i, j) refers to MixDis(i, j), and a(i, j) is the attribution message as defined below.

(2) The attribution message a(i, j).

The transmission from respective point lj to respective point li shows the attraction accumulation
of li to lj as the cluster center under the role of the respective points instead of lj , and the formula is
as follows: ÿ
api, jq “ mint0, rpj, jq ` maxt0, rpi1 , jquu (5)
i1 Rpi,jq

(3) The self-attribution message:

ÿ (
apj, jq “ max 0, rpj1 , jq (6)
j‰ j1

The passing of the above two kinds of messages between the respective points achieves the
division of center and the attaching point. Specifically, if j’ = i, reference point li is the center. Otherwise,
point lj’ is the center.
ISPRS Int. J. Geo-Inf. 2016, 5, 8 5 of 19

2.3. Positioning Point Set Searching

After the APC clustering, RP points will be divided into several classes according to different
signals and positions. When positioning, rough matching should be made between the MAC address
information received by the point to be positioned and the address vector of each class center to realize
the rough estimation of the class scope; more accurate matching can then be made by taking advantage
of the distance between the signal strength value of the point to be positioned and the signal strength
vector of each class center in particular.
(1) The class search space is filtered with the MAC address of the first N (3–5) APs corresponding
to the maximum signal strength RSSI received by the target in real time. As a result, one or more point
sets may be returned.
(2) The distance of the returned class center vector from the signal vector received in real time to the
previous step is calculated. The signal propagation characteristics show that the closer the geometric
distance between two data collection points, the more similar the signal strength values received. Thus,
the smallest class number of the signal vector distance is assigned to the final determined class.
(3) The search of the within-class fingerprint points adopts the Weighted K Nearest Neighbor
(WKNN) algorithm, which selects K nearest RPs to the signal domain of the point to be positioned and
takes the reciprocal of the distance as the weight to increase the weight of the position coordinates
of the fingerprint points with smaller distance to the signal domain, which can greatly improve the
positioning accuracy to some degree [30]. The average position coordinate of the RPs is calculated as
the positioning result of the users:

K
ř 1
p ˆ pxi , yi qq
di
px̂, ŷq “ i“1 (7)
řK 1

i “1 d i

where (xi , yi ) is the coordinate of the ith nearest RP in the determined fingerprint class collection,
px̂, ŷq is the positioning result, and di is the signal domain distance between the ith nearest RP and the
point to be positioned in the determined class. With respect to Equation (7), some other researchers
used 1{d2i instead of 1{di [31]. From the perspective of computational efficiency, the usage of 1{d2 will
further increase the impact of the signal distance on the calculation result. In other words, the ultimate
positioning result will tend to such an RP that has a smaller signal distance. However, because the
distance between two RPs in this paper is rather small, at 1.2 ˆ 1.2 m2 , such a correction will not
significantly affect the positioning result.

3. PDR and Wi-Fi Fusion Algorithm

3.1. Adaptive-Weighted Smoothing Filter Based on the Displacement Constraint

The Wi-Fi-based positioning system has an absolute positioning advantage, which usually updates
the position information every 1 or 2 s. The inertial measurement system can output high-rate original
observation data in real time, and the PDR positioning algorithm based on inertial data is featured
with strong independence, a short term, and high accuracy [32,33]. Owing to the finite velocity when
the pedestrian moves, it is inevitable that there is a connection between the present positioning result
and the previous positioning result in terms of physical distance. Based on this relation, first assume
that the distance threshold is δ S. When dead reckoning is made according to the data from the
inertial sensor, start from the previous point obtained through the fusion positioning to estimate
the current position information and calculate SPDR , the displacement within the time frame when
the wireless signal received by the terminal should vary. In addition, calculate SW IFI , the distance
between the current Wi-Fi-based positioning result and the previous position obtained through fusion
positioning. Then, adaptively choose the weight for the calculation made based on the integration of
ISPRS Int. J. Geo-Inf. 2016, 5, 8 6 of 19

both positioning technologies according to the relationship between the distance threshold δS and
|SW IFI ´ SPDR |, the absolute difference in the displacement distance separately through the Wi-Fi
and the PDR-based positioning method. When |SW IFI ´ SPDR | ă δS, the Wi-Fi-based positioning
technology with the absolute positioning characteristics is more weighted. In other words, choose
Formula (8) to estimate the current position of the pedestrian. Otherwise, return the positioning result
obtained in Formula (9) as the current position of the pedestrian.
$
maxpw1 , w2 qNk,w ` minpw1 , w2 qNk,p
& Nk “
’
’
w1 ` w2 (8)
’
’ E “ maxpw 1 , w 2 qE k,w ` minpw1 , w2 qEk,p
% k
w1 ` w2
$
minpw1 , w2 qNk,w ` maxpw1 , w2 qNk,p
& Nk “
’
’
w1 ` w2 (9)
’ minpw 1 , w 2 qEk,w ` maxpw1 , w2 qEk,p
% Ek “
’
w1 ` w2
` ˘
´ k, Nk,w
where pNk , Ek q represents the coordinate of the output from the fusion system at moment ¯ , Ek,w
is the coordinate of the output from the Wi-Fi positioning system at moment k, Nk,p , Ek,p is the
coordinate of the output from the PDR system at moment k, and pw1 , w2 q is the weight. According to
the experimental data, the measurement shows that the best positioning results can be achieved when
w1 and w2 are 0.4 and 0.6, respectively.

3.2. Adaptive System Noise Filter Based on the Pedestrian’s Moving Status
Based on the idea discussed in Section 3.1, to further optimize the weights of the two positioning
results, the EKF method is used for fusion. Actually, our fusion method has the advantage of being
able to perform a fusion calculation through the adaptive identification of the dynamic noise in the
filtering system according to the pedestrian’s moving status (walking forward, turning a corner, or
pseudo turning a corner) to realize a smaller trajectory divergence at the corner while improving
positioning accuracy.
First, judge the stationary or moving state according to the acceleration data in the inertial
measurement. In the initial positioning, gross error elimination is performed in the Wi-Fi positioning
result when pedestrians stand still for a few seconds. Then, calculate the average value as the starting
position of PDR. The fusion algorithm takes the position error, displacement distance difference and
course error as the state variable of the filter system:

X “ rdN, dE, ds, dθs (10)

where N and E represent the coordinates of the positioning system in the north and east, respectively,
obtained in real time; s is the step value calculated by the PDR algorithm in real time; and θ is the course
angle in the moving direction. When the Wi-Fi system position is updated, the position difference
between the Wi-Fi and PDR systems will be taken as the observed value of the system; namely,
” ıT
Z “ r∆N, ∆EsT “ Nw,k ´ Np,k , Ew,k ´ E p,k (11)

where p∆N, ∆Eq represents the position difference between the two
` ˘
´ systems
¯ at time point k, Nw,k , Ew,k
is the positioning result of the Wi-Fi system at time point k, and Np,k , E p,k is the position information
at time point k calculated according to the dead reckoning principle.
ISPRS Int. J. Geo-Inf. 2016, 5, 8 7 of 19

The state equation of the EKF system is

$
’
’ dNk`1 “ dNk ` cosθk ˆ dsk ´ sk ˆ sinθk ˆ dθk ` dw N
’
& dE
k`1 “ dEk ` sinθk ˆ dsk ` sk ˆ cosθk ˆ dθk ` dw E
(12)
’ dsk`1 “ dsk ` dws
’
’
dθk`1 “ dθk ` dwθ
%

here the dynamic noise satisfaction of the position coordinates and displacement meets the
` ˘ ` ˘ ` ˘
Gaussian distribution—namely, w N „ N 0, δ2N , wE „ N 0, δE2 , ws „ N 0, δs2 .
State-transition matrix:
» fi
1 0 cosθk ´sk ˆ sinθk
— 0 1 sinθ
k sk ˆ cosθk ffi
Φk “ — (13)
— ffi
ffi
– 0 0 1 0 fl
0 0 0 1

Measurement matrix: « ff
1 0 0 0
Hk “ (14)
0 1 0 0

For the dynamic noise sequence and measurement noise sequence required in the filtering process,
many positioning result data are used to obtain the experience numbers via characteristic analysis of
the error statistics of the Wi-Fi and PDR positioning systems. The dynamic noise matrix is:
» fi
δ2N 0 0 0
— 0 δE2 0 0 ffi
Qk “ — (15)
— ffi
δs2
ffi
– 0 0 0 fl
0 0 0 δθ2

Given δ2N “ δE2 “ 2, δs2 “ 1, measurement noise matrix:

« ff
Rw 0 0 0
R“ (16)
0 Rw 0 0

where Rw “ 102 . Considering the actual situation in which the value of the course wθ is larger when
turning and smaller when moving straight, adaptively determine the dynamic noise value of the
course-related variables by judging the pedestrian movement attributes (straight or turning). Use the
accumulated value SGyr (rad/s) of the gyro data in each step to determine the movement attributes of
mobile users. The judgement conditions of turning are

SGyrk ˆ SGyrk´1 ą 0 and abspSGyrk q ą 10rad{s (17)

In the experiment, we found that the results will improve if we dynamically adjust the Q matrix
each time there are turning trends. The turning trends contain the true turning and pseudo turning,
which is caused by the large deviation that occurs when a pedestrian holds the phone in a certain
direction while walking. The formula for judging the turning tread is:

SGyrk ˆ SGyrk´1 ą 0 (18)

´ ˝ ¯2 ´ ˝ ¯2
When the pedestrian moves straight, δθ2 “ 2 ; when the pedestrian turns, δθ2 “ 15 ; when
´ ˝ ¯2
the pedestrian moves straight but pseudo turning is caused by a slight turning of the hand, δθ2 “ 8 .
ISPRS Int. J. Geo-Inf. 2016, 5, 8 8 of 19

ISPRSIn
Int.the case in
J. Geo-Inf. which
2016, the Wi-Fi position
has not been updated, take the difference between
5, 8; doi:10.3390/ijgi5010008 8 ofthe
19
coordinates predicted by the system and the coordinates observed through the PDR method as the
observation variables
variablesto
toperform
performa arecursive
recursive correction
correction on on
thethe calculation
calculation of position
of the the position through
through PDR
PDR with
with all all others
others remaining
remaining unchanged:
unchanged:

ZZ“[N , ∆Es
r∆N, E]TT “rN
[ Nk`k 11´NNp,kp`, k 1,1E, E  E T]T
k`k11´ E p,kp`, k1s1
(19)
(19)

where the coordinates

where the coordinates predicted
predicted by
by the
the system
system can
can be
be updated
updated through
through the
the following
following approach:
approach:

k 1  “
NN NN p , k ` dN  ( s  ds )  cos(  d )
dN ` pskk ` dskk q ˆ cospθkk` dθk qk
k `1 p,k
(20)
(20)
EkE1k`1 E p , k  dE  ( sk  dsk )  sin( k  d k )
“ E p,k ` dE ` psk ` dsk q ˆ sinpθk ` dθk q

According to
According tothe
theposition
positionerror
errorobtained
obtainedthrough
through the
the filtering,
filtering, Formula
Formula (20)(20) is able
is able to update
to update the
the pedestrian’s current position, which is also the final position at moment k + 1 obtained
pedestrian’s current position, which is also the final position at moment k + 1 obtained after after the
the
computation is
computation is performed
performed inin the
the fusion
fusion model.
model.

4. Experimental Analysis
4. Experimental Analysis
The experimental site was set up on the fourth floor of the School of Environment Science and
The experimental site was set up on the fourth floor of the School of Environment Science and
Spatial Informatics, China Mining University of Technology. A Wi-Fi network system was established,
Spatial Informatics, China Mining University of Technology. A Wi-Fi network system was
and the Samsung Galaxy Note 3 (SM-N9002) was chosen as the mobile testing device. As shown
established, and the Samsung Galaxy Note 3 (SM-N9002) was chosen as the mobile testing device. As
in Figure 1, D-LINK wireless routers (DIR-600NB) were installed every 9 m along both sides of
shown in Figure 1, D-LINK wireless routers (DIR-600NB) were installed every 9 m along both sides
the corridor, providing 64 APs in total. In the positioning process, the positioning system sent the
of the corridor, providing 64 APs in total. In the positioning process, the positioning system sent the
positional information to mobile users every 2 s according to the Wi-Fi signal strength. The data
positional information to mobile users every 2 s according to the Wi-Fi signal strength. The data
sampling frequency of the inertial sensor was 50 Hz; each new data point was analyzed, including the
sampling frequency of the inertial sensor was 50 Hz; each new data point was analyzed, including
gait detection, step-size calculation, course rough calculation, and position calculation. The position
the gait detection, step-size calculation, course rough calculation, and position calculation. The
was calculated according to the fusion model described in Section 3 with the Wi-Fi positioning results
position was calculated according to the fusion model described in Section 3 with the Wi-Fi
and PDR positioning results. In the experiment, the pedestrian started from the west end of Area C in
positioning results and PDR positioning results. In the experiment, the pedestrian started from the
the School of Environment Science and Spatial Informatics towards the westernmost end of Area B via
west end of Area C in the School of Environment Science and Spatial Informatics towards the
Area A along the corridor at a constant speed. In this process, the pedestrian basically held the mobile
westernmost end of Area B via Area A along the corridor at a constant speed. In this process, the
phone level and walked 316 steps in total.
pedestrian basically held the mobile phone level and walked 316 steps in total.

(a) (b)
Figure 1. Experimental site. (a) Floor scene graph of the fourth floor; (b) 3D model of the experimental
Figure 1. Experimental site. (a) Floor scene graph of the fourth floor; (b) 3D model of the experimental site.
site.

4.1. Wi-Fi Positioning Analysis

4.1. Wi-Fi Positioning Analysis
The experimental site was set up in a long, narrow corridor environment divided into A, B and C
The experimental site was set up in a long, narrow corridor environment divided into A, B and
zones, and a regular quadrilateral (1.2 m ˆ 1.2 m) grid structure was used to lay out the fingerprint
C zones, and a regular quadrilateral (1.2 m × 1.2 m) grid structure was used to lay out the fingerprint
points to achieve offline fingerprint data acquisition and construct the fingerprint database. The
points to achieve offline fingerprint data acquisition and construct the fingerprint database. The
sampling rate of the device to the received signal is 100 Hz, and relevant data collected by the RPs
usually includes Wi-Fi signal strength and the real coordinates of RPs. The received signal strength
ISPRS Int. J. Geo-Inf. 2016, 5, 8 9 of 19

ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008 9 of 19

ISPRS Int. J. rate
sampling Geo-Inf.
of 2016, 5, 8; doi:10.3390/ijgi5010008
the device signal is 100 Hz, and relevant data collected by the9 RPs
to the received of 19

usually
is includes Wi-Fi signal strength and the real coordinates of RPs. The received signal strengthpoint
is an
is an
an integer
integer in
in (−110,
(−110, −15);
−15); 390
390 efficient
efficient points
points are
are finally
finally collected
collected from
from the
the fingerprint
fingerprint point
integer in (´110,
distribution, ´15); 390Figure
efficient points are finally collected from the fingerprint point distribution,
distribution,as
asshown
shownin in Figure22[34].
[34].
as shown in Figure 2 [34].

BB

CC
A
A

Figure
Figure2.
Figure 2.Distribution
2. Distributionof
Distribution offingerprint
of fingerprintpoints.
fingerprint points.
points.

(1)
(1)Offline
(1) Offlinedata
Offline dataacquisition
data acquisitionand
acquisition andpreprocessing
and preprocessing
preprocessing
The
TheWi-Fi
The Wi-Fisignal
Wi-Fi signalfrom
signal fromthe
from theD-LINK
the D-LINK(DIR-600NB)
D-LINK (DIR-600NB)wireless
(DIR-600NB) wirelessrouter
wireless routerwas
router wasrecorded
was recordedin
recorded inaaafixed
in fixedposition
fixed position
position
on
on the fourth floor for five days to describe the time-variation effect of the Wi-Fi signal with the
on the
the fourth
fourth floor
floor for
for five
five days
days to
to describe
describe the
the time-variation
time-variation effect
effect of
of the
the Wi-Fi
Wi-Fi signal
signal with
with the
the
measured
measured data.
measured data. The
data. The statistical
The statistical results
statistical results are
results are shown
are shown in
shown in Figure
in Figure 3.
Figure3.
3.

Figure 3. The polar coordinates diagram for statistical results of the Wi-Fi signal in the five
Figure 33．
Figure ．The
consecutive The polar coordinates diagram for statistical results of the Wi-Fi signal in the five
days.polar coordinates diagram for statistical results of the Wi-Fi signal in the five
consecutive
consecutivedays.
days.

From the
From figure, wewe can seesee that the signal propagation was stable from 23:00 to 7:00 buttofluctuated
From the the figure,
figure, we can can see thatthat the
the signal
signal propagation
propagation was was stable
stable from
from 23:00
23:00 to 7:00
7:00 but
but
greatly
fluctuatedfrom 7:00 to 23:00. In fact, there were staff members in the lab beginning at 7:00, andat the lab
fluctuated greatly from 7:00 to 23:00. In fact, there were staff members in the lab beginning at 7:00,
greatly from 7:00 to 23:00. In fact, there were staff members in the lab beginning 7:00,
closed
and at 23:00. Thus, we can see that the signal is significantly affected by human disturbance. In
and the
the lab
lab closed
closed at at 23:00.
23:00. Thus,
Thus, we we can
can see
see that
that the
the signal
signal isis significantly
significantly affected
affected by by human
human
addition,
disturbance.by comparing the data from the five days, we can find that the RSSI of the curves at the same
disturbance. In In addition,
addition, by by comparing
comparing the the data
data from
from the the five
five days,
days, wewe can
can find
find that
that the
the RSSI
RSSI ofof the
the
time every
curves day are different bydayup to 15different
dB. In Wi-Fi fingerprint positioning, the fingerprint collection
curves at the same time every day are different by up to 15 dB. In Wi-Fi fingerprint positioning, the
at the same time every are by up to 15 dB. In Wi-Fi fingerprint positioning, the
is distributed
fingerprint based onisthe current signal strength; the signalsignal
changes at different times in the same
fingerprint collection is distributed based on the current signal strength; the signal changes at
collection distributed based on the current strength; the signal changes at
day, and changes
different in thesame
same period on differentindays will result in a failure to completely describeinthe
different times
times in in the
the same day, day, and
and changes
changes in the the same
same period
period on on different
different days
days will
will result
result in aa
signal strength
failure distribution bythe the statistics of the signal strength collected in onethe period, which will
failure to
to completely
completely describe
describe the signal
signal strength
strength distribution
distribution by by the
the statistics
statistics ofof the signal
signal strength
strength
undoubtedly
collected affect the positioning results.
collectedin inone
oneperiod,
period,which
whichwill willundoubtedly
undoubtedlyaffect affectthe thepositioning
positioningresults.
results.
Therefore, we
Therefore, we collected
collected fingerprint
fingerprint datadata on on 10
10 consecutive
consecutive days days from 11 11 to 2020 April 2015.
2015. Ten
Therefore, we collected fingerprint data on 10 consecutive days from from 11 to to 20 April
April 2015. TenTen
samples were
samples were collected
collected ininfour
four directions
directionsby each
by RP each
each RP day, resulting
each day, in 40 groups
resulting in 40 of signalofstrength
groups signal
samples were collected in four directions by each RP each day, resulting in 40 groups of signal
strength
strength measurements
measurements in in total.
total. Figure
Figure 44 shows
shows thethe signal
signal strength
strength information
information of of 10
10 APs
APs observed
observed
on
on an RP, and all 40 signals of each AP are very unstable. The data were collected from fourdirections
an RP, and all 40 signals of each AP are very unstable. The data were collected from four directions
because
becausethere
therearearelarge
largedifferences
differencesin inthe
thesignal
signalstrength
strengthvaluesvaluesobtained
obtainedin indifferent
differentdirections
directionsat atthe
the
ISPRS Int. J. Geo-Inf. 2016, 5, 8 10 of 19

measurements in total. Figure 4 shows the signal strength information of 10 APs observed on an RP,
and
ISPRSall
Int.40 signals2016,
J. Geo-Inf. of each AP are very unstable. The data were collected from four directions because
5, 8; doi:10.3390/ijgi5010008
ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008
10 of 19
10 of 19
there are large differences in the signal strength values obtained in different directions at the same
same position.
position.
position.
same Figure
FigureFigure
5 shows 55 shows
shows the signal
the signal
the signal
strengthstrength distribution
distribution
strength ofsame
of theof
distribution the same
the same AP collected
collected
AP collected
AP by
by anby aninRP
RPan RP in
four
in
four
directionsdirections
(absolute (absolute
value), where value),
the where
maximum the
difference
four directions (absolute value), where the maximum difference is up to maximum
is up to 20 difference
dBm. The is
measurementup to
of
20 dBm.
difference
20 dBm. TheThe measurement
is affected
measurement of difference
by theofmultipath is
differenceeffect affected by
causedby
is affected the
bythe multipath
themultipath effect
layout ofeffect caused by
the environment. the layout
caused by theMoreover, of the
layout of the
environment.
body itself is Moreover,
one of the the body
important itself is
factors one of the
interferingimportant
with factors
wireless interfering
signal
environment. Moreover, the body itself is one of the important factors interfering with wireless with
propagation. wireless signal
Finally, the
signal
propagation.
mean Finally,
value ofFinally,
propagation. the
all fieldthe mean
strengths value of all
in theoffour
mean value field strengths
directions
all field in
collected
strengths the four
in theover directions
four10 collected
days is taken
directions over
as the
collected 10 days
fingerprint
over 10 days
is taken
taken as
information
is as the
the fingerprint
of fingerprint
the RP. information of
information of the
the RP.
RP.

Figure 4.
Figure 4. Signal strength
strength data of
of 10 available
available APs for
for an RP.
RP.
Figure 4. Signal
Signal strength data
data of 10
10 available APs
APs for an
an RP.

78.5667 dBm
78.5667 dBm
78.5667 dBm
78.5667 dBm

65.7
65.7 dBm
65.7 dBm
dBm 75.4 dBm
75.4 dBm
75.4 dBm
65.7 dBm 75.4 dBm
RP
RP

67.5 dBm
67.5 dBm
67.5 dBm
67.5 dBm

Figure 5. Signal strength observed from four sampling directions.

Figure 5.
Figure 5. Signal
Signal strength
strength observed
observed from
from four
four sampling
sampling directions.
directions.

(2)
(2) Clustering
Clustering analysis
analysis
(2) Clustering analysis
To
To show
showthe theanalysis
analysisresults,
results,the theexperiment
experiment re-collected
re-collected 93 efficient points
93 efficient
efficient in Areas
points C and
in Areas
Areas CA with
and A
an To
average show the
sampling analysis
intervalresults,
of 3.5 the
m; experiment
the APC re-collected
algorithm is 93
adopted to clusterpoints
the in
respective C and
points A
in
with an
with an average
average sampling
sampling interval
interval of of 3.5
3.5 m;
m; the
the APC
APC algorithm
algorithm is is adopted
adopted to to cluster
cluster the
the respective
respective
the sample
points in thedatabase
sample by extracting
database by the fusion
extracting feature.
the fusion feature.
pointsBy incomparing
the sample database
Figures 6 6andby7extracting
it7,can the fusion
be found thatthatfeature.
the results obtained by APC
By
By comparing
comparing Figures
Figures 6 and
and 7, it
it can
can be
be found
found that the results
the results obtained
obtained by clustering
by APC
are closer
APC clustering
clustering are
are
to the geographic
closer toto the division,
the geographic such
geographic division, as
division, suchthe corner
such as as the of the
the corner corridor
corner of of the and the
the corridor open
corridor and area,
and the and
the open the classification
open area,
area, and
and the
the
closer
results are more uniform. This is becauseThis the clustering feature in the APC method fused part of the
classification
classification results are more
results are Although uniform.
more uniform. is
Thisdistancebecause
is because the
the clustering
clustering feature in the APC method
geographic
fused part information.
of the geographic the center
information. Although hascenter
the increased, the feature
distance distance
has
incontrol
the APC
increased, later
the
method
has an
distance
fused part
important of
role the
inangeographic
theimportant
improvement information.
of the Although
the improvement the
positioning accuracy. center distance has increased, the distance
control later
control later has
has an important role in
role in the improvement of the
of the positioning
positioning accuracy.
accuracy.
The selection of the K-means clustering initial
The selection of the K-means clustering initial value is random, so the value is random, so the results
results generated
generated by by
several operation
several operation iterations
iterations are
are different.
different. Because
Because the the APC
APC algorithm
algorithm is is an
an unsupervised
unsupervised classification
classification
method, it is unnecessary to set the number of classes and the random
method, it is unnecessary to set the number of classes and the random initial value, which is initial value, which is the
the result
result
of long-term feedback. Therefore, the obtained calculation
of long-term feedback. Therefore, the obtained calculation result is unique. The K-means result is unique. The K-means
ISPRS Int. J. Geo-Inf. 2016, 5, 8 11 of 19

The selection of the K-means clustering initial value is random, so the results generated by several
operation iterations
ISPRS Int. J. Geo-Inf. 2016,are
5, 8;different. Because the APC algorithm is an unsupervised classification method,
doi:10.3390/ijgi5010008 11 of 19
ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008 11 of 19
it is unnecessary to set the number of classes and the random initial value, which is the result of
classification results
long-term feedback.
classification are
are shown
results Therefore, shown thein Figure
inobtained 7.
7. The
The circles
Figure calculation resultare
circles the
the positions
is unique.
are positions with
with unreasonable
The K-means classification
unreasonable
classification:
classification: 1 is due to few class members; 2 is because of fuzzy boundaries, and 3 is because
results are 1
shown is due
in to
Figure few7. class
The members;
circles are 2
the is because
positions of
withfuzzy boundaries,
unreasonable and 3 is
classification: 1 is duethe
because to
the
corner
corner of
of the
few class the corridor
members;
corridor 2 is not
not classified.
is because of fuzzy
classified. Table 11 shows
shows the
boundaries,
Table andresult
the statistics
3 is because
result theafter
statistics the
corner
after K-means
theof clustering
the corridor
K-means is not
clustering
algorithm
classified. was
Tablerun1 10
shows
algorithm was run 10 times. times.the result statistics after the K-means clustering algorithm was run 10 times.

Figure
Figure 6.
6. APC
APCclassification
classification results.
results.
Figure 6. APC classification results.

Figure 7. Classification results of K-means.

Figure
Figure 7.
7. Classification
Classification results
results of
of K-means.
K-means.

Table 1. The statistics for the number of class members in the result of K-means clustering (10 times).
Table
Table 1.
1. The
The statistics
statistics for
for the
the number
number of
of class
class members
members in
in the
the result
result of
of K-means
K-means clustering
clustering (10
(10 times).
times).
Times Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Class 8 Class 9
Times Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Class 8 Class 9
Times11 Class171
17 Class15 2
15
10 3
Class
10
7 4
Class
7
6 5
Class
6
8
Class
8 6 10
Class 7 11
10
9
11Class 8 9 Class 9
2 17 15 10 7 9 15 8 3 9
1 2 1717 1515 10
10 77 96 15 8 8 10 3 11 9 9
3
2 3 17
1717 1566 10
10 127 10 8 10 11 9 9
10 12 109 8 15 10 8 11 3 9
3 44 1711
11 666 106
6
10
12
10
910
9
7 8
7
9 10
9
15 11
15
20 9
20
4 55 11 88 699 666 10
10
10
129
12
10 7
10
8 9
8
10 15
10
20 20
20
5 6 8 17 96 10
6 9
10 712 9 10 15 8 11 10 9 20
6 17 6 10 9 7 9 15 11 9
6 7 1717 66 10
10 99 77 6 9 8 15 10 11 20 9
7 17 6 10 9 7 6 8 10 20
7 8 17 9 69 10
14 109 77 9 6 15 8 11 10 9 20
8 98 9 89 999 14
146
10
10
10
7
127
9
10 9
15
18 15
11
11 11
9
9 9
9 9 88 99 66 10
10 12 1010 18 18 11 11 9
10 32 13 10 8 1012 8 3 5 4 9
10 10 3232 1313 10
10 88 1010 88 3 3 5 5 4 4

It
It can
can be
be seen
seen from
from Table
Table 11 that
that there
there are
are great
great differences
differences between
between thethe classification
classification results
results
underIt the
can be seenof
from Table 1initial
that there are great differences between the classification results using
under
under the action of different initial values, which illustrates that instability results
action different values, which illustrates that instability results when
when using
the action
K-means of different initial values, which illustrates that instability results when using K-means to
K-means to to cluster
cluster the
the respective
respective points
points in
in the
the fingerprint
fingerprint database.
database. Moreover,
Moreover, thethe classification
classification
number
number should
should bebe set
set before
before processing,
processing, which
which isis more
more difficult
difficult for
for Wi-Fi
Wi-Fi signals,
signals, which
which are
are
susceptible to interference.
susceptible to interference.
(3)
(3) Analysis
Analysis results
results of
of static
static positioning
positioning
Static
Static positioning
positioning test
test method:
method: take
take 26
26 points
points to
to be
be positioned
positioned inin the
the experimental
experimental site;
site; sample
sample
for
for 10 s at the points to be positioned in the up, down, left, and right directions with the same device;
10 s at the points to be positioned in the up, down, left, and right directions with the same device;
 ISPRS Int. J. Geo-Inf. 2016, 5, 8 12 of 19

cluster the respective points in the fingerprint database. Moreover, the classification number should be
set before processing, which is more difficult for Wi-Fi signals, which are susceptible to interference.

(3) Analysis results of static positioning

ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008 12 of 19
Static positioning test method: take 26 points to be positioned in the experimental site; sample
for 10
and records at the
the points
mean to be positioned
value in thesignal
of the Wi-Fi up, down,
data.left,
Theand right directions
experiment withand
analyzes the same device;the
compares
and recordeffects
positioning the mean
of APC value
andofK-means
the Wi-Fi signal
from thedata.
aspectsTheof experiment analyzes
time efficiency and compares
and positioning the
accuracy.
positioning effects of APC and K-means from the aspects of time efficiency and positioning
The positioning times of the 26 points to be positioned are shown in Figure 8. After the clustering accuracy.
The positioning times of the 26 points to be positioned are shown in Figure 8. After the clustering
process, both the data search volume and positioning time are reduced. Compared with a global
process, both the data search volume and positioning time are reduced. Compared with a global
search, K-means and APC clustering processing achieve great improvement in time efficiency. The
search, K-means and APC clustering processing achieve great improvement in time efficiency. The
average positioning time of global search is 127 ms, and K-means and APC have similar effects, which
average positioning time of global search is 127 ms, and K-means and APC have similar effects, which
are 76 ms and 70 ms, respectively. The largest drop by K-means is 88.2%, and the smallest drop is
are 76 ms and 70 ms, respectively. The largest drop by K-means is 88.2%, and the smallest drop is
only 0.2%, whereas the largest drop by APC is 87.2%, and the smallest drop is up to 16.8%. The
only 0.2%, whereas the largest drop by APC is 87.2%, and the smallest drop is up to 16.8%. The
quantity
quantityofofdata
datasearched
searchedby byAPC
APCdecreases
decreases by 75% on
by 75% on average
averageand and80%
80%atatmost.
most.Because
Because the
the average
average
reduction
reduction rates of APC and K-means are almost the same, the average positioning time of thethe
rates of APC and K-means are almost the same, the average positioning time of twotwo
clustering
clusteringmethods
methodsisissimilar;
similar;however,
however, K-means
K-means has has aa larger
largerdrop
dropbecause
becauseofofthe
thegreat
great difference
difference in in
thethe
numbers
numbers of sampling points searched on each positioning point caused by the non-uniformity of of
of sampling points searched on each positioning point caused by the non-uniformity
thethe
clustering
clusteringresults.
results.

Comparison of static
Figure8.8.Comparison
Figure static positioning
positioningtime.
time.

AsAsshown
shownininFigure
Figure9,9,the
theclustered
clustered positioning result isismore
positioning result moreaccurate
accuratethan
thanthat
thatofof
thethe global
global
search. The average positioning accuracy of the global search is 2.6 m, but the positioning
search. The average positioning accuracy of the global search is 2.6 m, but the positioning accuracies accuracies
of of
K-means
K-meansand andAPC
APCareare 1.6
1.6 m and
and 1.5
1.5m,
m,respectively;
respectively;the the maximum
maximum positioning
positioning errorerror of K-means
of K-means is
is 5.6 m, and the minimum positioning error is 0.18 m; the maximum positioning error of APC is
5.6 m, and the minimum positioning error is 0.18 m; the maximum positioning error of APC is 2.8 m,
2.8and
m, the
andminimum
the minimumpositioning error is
positioning 0.2 m.
error In terms
is 0.2 m. In of the average
terms error, APC
of the average hasAPC
error, a slightly
has a better
slightly
result than K-means and has a smaller error fluctuation. The class with few members
better result than K-means and has a smaller error fluctuation. The class with few members in the in the K-means,
combined
K-means, with the fact
combined withthat
thethe mean
fact thatvalue of the signal
the mean value ofofthat
the class
signalcannot reflect
of that classthecannot
features of thisthe
reflect
kind of area signal, can easily result in mismatching and thus reduce the accuracy.
features of this kind of area signal, can easily result in mismatching and thus reduce the accuracy.
ISPRS Int. J. Geo-Inf. 2016, 5, 8 13 of 19
ISPRS
ISPRS Int.
Int. J.
J. Geo-Inf.
Geo-Inf. 2016,
2016, 5,
5, 8;
8; doi:10.3390/ijgi5010008
doi:10.3390/ijgi5010008 13
13 of
of 19
19

Figure 9. Comparison of static positioning accuracy.

accuracy.
Figure 9. Comparison of static positioning accuracy.

(4) Triangle
(4) Triangle mesh
Triangle mesh structure
meshstructure
structureof of fingerprint
offingerprint points
fingerprintpoints
points
(4)
To further
To further improve
further improve
improve the the computation
the computation efficiency,
computation efficiency,
efficiency, wewe attempted
attempted aaa triangle
we attempted triangle mesh
triangle mesh structure
mesh structure layout
structure layout
layout
To
of fingerprint
of fingerprint points,
points, as as shown
shown in in Figure
Figure 10.10. There
There is is no
no significant
significant difference
difference in in the
the positioning
positioning
of fingerprint points, as shown in Figure 10. There is no significant difference in the positioning
accuracies
accuracies of of the
of the APC
the APC clustering
APC clustering
clustering of of the
of the fingerprint
the fingerprint database
fingerprint database based
database based
based onon the
on the two
the two structural
two structural layouts.
structural layouts. The
layouts. The
The
accuracies
static
static positioning
positioning accuracy
accuracy is
is within
within 2
2 m
m on
on average,
average, and
and the
the average
average accuracy
accuracy of
of dynamic
dynamic positioning
positioning
static positioning accuracy is within 2 m on average, and the average accuracy of dynamic positioning
is approximately
is approximately 4 m, m, asas shown
shown in in Table
Table 2.2. However,
However, considering
considering thethe workload
workload of of the
the off
off sampling,
sampling,
is approximately 44 m, as shown in Table 2. However, considering the workload of the off sampling,
the triangular
the triangular mesh
triangular mesh
mesh of of fingerprint
of fingerprint points
fingerprint points is
points is half
is half the
half the square
the square grid
square grid in
grid in accordance
in accordance
accordance withwith the
with the layout
the layout plan
layout plan
plan ofof
of
the
this
this paper;
paper; that
that is,
is, compared
compared with
with the
the quadrangular
quadrangular layout
layout plan,
plan, the
the workload
workload of
of the
the offline
offline sampling
sampling
this paper; that is, compared with the quadrangular layout plan, the workload of the offline sampling
is reduced
is reduced
reduced by by half,
by half, and
half, and the
and the two
the two layout
two layout plans
layout plans contain
contain 195
plans contain 195 and 390 fingerprint points, respectively.
is 195 and
and 390
390 fingerprint
fingerprint points,
points, respectively.
respectively.
Therefore,
Therefore, the
the triangular
triangular mesh
mesh fingerprint
fingerprint reference
reference points
points can
can be
be set
set in
in the
the environment
environment with long
with long
long
Therefore, the triangular mesh fingerprint reference points can be set in the environment with
and narrow
and narrow features
features forfor the
the collection
collection of of Wi-Fi
Wi-Fi fingerprint
fingerprint data.
data.
and narrow features for the collection of Wi-Fi fingerprint data.
The inconsistency
The inconsistency
inconsistency of of dynamic
of dynamic
dynamic and and static
and static positioning
static positioning accuracy
positioning accuracy
accuracy isis caused
is caused
caused by by the
by the static
the static positioning
static positioning
positioning
The
data,
data, consisting
consisting of
of the
the mean
mean value
value of
of multi-data
multi-data collected
collected from
from four
four directions,
directions, so
so the
the positioning
positioning
data, consisting of the mean value of multi-data collected from four directions, so the positioning
accuracy is
accuracy is higher;
higher; however,
however, the the Wi-Fi
Wi-Fi dynamic
dynamic positioning
positioning datadata are
are collected
collected byby the
the device
device inin real
real
accuracy is higher; however, the Wi-Fi dynamic positioning data are collected by the device in real
time, which
time, which only
which only uses
only uses a group
uses aa group
group of of scanning
of scanning data.
scanning data. That
data. That is,
That is, when
is, when
when thethe signal
the signal scanned
signal scanned
scanned by by the
by the device
the device
device isis
is
time,
changed
changed or
or updated,
updated, the
the new
new data
data are
are used
used immediately
immediately for
for the
the positioning
positioning calculation.
calculation. Owing
Owing to
to
changed or updated, the new data are used immediately for the positioning calculation. Owing to
the randomness
the randomness
randomness and and the
and the instability
the instability
instability of of the
of the signal,
the signal, the
signal, the positioning
the positioning accuracy
positioning accuracy
accuracy isis lower
is lower than
lower than the
than the static
the static data
static data
data
the
position
position calculation.
calculation.
position calculation.

Figure 10. Triangular mesh structure layout of fingerprint points.

Figure 10. Triangular mesh structure
structure layout
layout of
of fingerprint
fingerprint points.
points.

Table
Table 2.
2. Positioning
Positioning results
results based
based on
on different
different layout
layout fingerprint
fingerprint databases.
databases.
ISPRS Int. J. Geo-Inf. 2016, 5, 8 14 of 19

Table 2. Positioning results based on different layout fingerprint databases.

Quadrilateral fingerprint database Triangle fingerprint database

Average error of static positioning / m 1.50 1.76
Maximum error of static positioning / m 2.80 3.52
Average error of dynamic positioning / m 4.09 4.43
Maximum error of dynamic positioning / m 19.76 22.4

4.2. Fusion Analysis

Because the Wi-Fi positioning system has an absolute positioning function, the positioning
ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008
results
14 of 19
of the system can be considered as the PDR initial point. By calculation, the initial position error of the
quadrilateral fingerprint database positioning is 2.58 m, which uses four kinds
Quadrilateral of positioning schemes
Triangle
for the indoor positioning test: fingerprint database fingerprint database
Scheme 1: Using the WKNN
Average error of static positioning / m positioning model based
1.50 on a Wi-Fi signal strength affinity clustering
1.76
fingerprint database to mark Wi-Fi.
Maximum error of static positioning / m 2.80 3.52
Scheme
Average 2: error
Withofthe Wi-Fi static
dynamic smooth
positioning / m positioning
4.09 results as the starting
4.43 point, mark the PDR
positioning based on compensation course
Maximum error of dynamic positioning / m 19.76 reckoning as PDR. Here, the estimation
22.4 of course error is
based on the method proposed by Chen in 2010 [35].
Scheme
4.2. Fusion3: Analysis
According to the Wi-Fi positioning and course correction PDR, mark WEPDR based on
conventional EKF
Because the value fusion
Wi-Fi positioning.
positioning system has an absolute positioning function, the positioning
Scheme
results of4:theAccording
system can to be
theconsidered
Wi-Fi positioning
as the PDRand course
initial correction
point. PDR, mark
By calculation, AWEPDR
the initial positionbased
on theerror
fusion
of thepositioning
quadrilateral of the adaptive database
fingerprint model noise extended
positioning Kalman
is 2.58 filter. uses four kinds of
m, which
The trajectory
positioning of positioning
schemes for the indoor results obtained
positioning test:by the different positioning models is shown in
Figure 11. Scheme 1: Using the WKNN positioning model based on a Wi-Fi signal strength affinity
clustering fingerprint database to mark Wi-Fi.
For the user trajectory of the Wi-Fi positioning system in Scheme 1, two apparent instabilities are
shown in Scheme
the figure: 2: With
first,the Wi-Fi
there arestatic
many smooth positioning
occasions results as phenomenon
of the rebound the starting point, mark thetoPDR
(according the user
positioning based on compensation course reckoning as PDR. Here, the estimation of course error is
moving direction in this experiment, the normal movement trajectories can be only forward along the
based on the method proposed by Chen in 2010 [35].
route). The concept of “rebound” in this paper characterizes the received rear position information
Scheme 3: According to the Wi-Fi positioning and course correction PDR, mark WEPDR based
fed back by the system
on conventional EKFinvalue
the forward positioning; second, there are repeated accumulation points in
fusion positioning.
positioning. That is, the user received
Scheme 4: According to the Wi-Fi positioning feedback of the
andsame
courseposition orPDR,
correction a nearby
mark position
AWEPDRinbased
different
positions at different time points, which caused the clustering
on the fusion positioning of the adaptive model noise extended Kalman filter. of positioning points. It is found that
those twoThe issues are a large
trajectory problemresults
of positioning independently
obtained bybased on the positioning
the different Wi-Fi signal fingerprint
models is shown database
in
Figure 11.
positioning algorithm.

(a)

Figure 11. Cont.

ISPRS Int. J. Geo-Inf. 2016, 5, 8 15 of 19
ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008 15 of 19

(b)

Figure 11. Trajectory analysis for different indoor positioning methods. (a) Trajectory of PDR,
Figure 11. Trajectory analysis for different indoor positioning methods. (a) Trajectory of PDR,
WIFI,REAL; (b) trajectory of AEKFPDR, EKFPDR.
WIFI,REAL; (b) trajectory of AEKFPDR, EKFPDR.
For the user trajectory of the Wi-Fi positioning system in Scheme 1, two apparent instabilities
In
areScheme
shown in2,the owing
figure:tofirst,
thethere
errorare ofmany
gait detection,
occasions ofstep-size
the reboundcalculation,
phenomenon and(according
course estimation,
to the
thereuser
are some
movingerrors withinthe
direction thisPDR positioning
experiment, results.
the normal Owing trajectories
movement to the error canofbethe initial
only position
forward
along the route). The concept of “rebound” in this paper characterizes
obtained by Wi-Fi positioning, the position of the overbridge from Area B to Area A especially deviates the received rear position
from information
the structures. fed However,
back by the system in the
its advantage forward
is that positioning;
the output second,
positioning thereare
points areuninterrupted,
repeated
accumulation
and there is no rebound points or in clustering
positioning.phenomenon.
That is, the user received feedback of the same position or a
nearby position in different positions at different time points, which caused the clustering of
In Scheme 3, the point positions are evenly distributed, and there is little rebound, which alleviates
positioning points. It is found that those two issues are a large problem independently based on the
the pileup problems to some degree. It can be seen that the classic EKF fusion algorithm preserves
Wi-Fi signal fingerprint database positioning algorithm.
the continuity and short-term
In Scheme 2, owing to stability
the error of of the
gait pedestrian dead reckoning,
detection, step-size calculation,which greatly
and course weakens the
estimation,
error there
accumulation of the PDR and has a certain capability to resist the
are some errors with the PDR positioning results. Owing to the error of the initial position gross Wi-Fi error. With the
increasing fusion times, the deviated points will gradually return to the
obtained by Wi-Fi positioning, the position of the overbridge from Area B to Area A especially real track trajectory direction.
Although the initial
deviates from the positioning
structures. error is approximately
However, its advantage 4.2 m, the positioning
is that accuracy can
the output positioning be gradually
points are
uninterrupted,
compensated to some anddegree
there iswith
no rebound
the update or clustering
of Wi-Fiphenomenon.
positions. However, the divergence problem of
In Scheme
point positions 3, the at
still exists point positions are evenly distributed, and there is little rebound, which
the corners.
In Scheme 4, the noise matrixtovalue
alleviates the pileup problems some Q degree. It can be seen
is dynamically set that the classic
according EKFpedestrian
to the fusion algorithm
movement
preserves the continuity and short-term stability of the pedestrian dead reckoning, which greatly
attributes (straight or turning). Figure 12 shows the accumulated gyro value corresponding to each step of
weakens the error accumulation of the PDR and has a certain capability to resist the gross Wi-Fi error.
the pedestrian, in which the singular points in the three rectangular boxes represent the three occasions of
With the increasing fusion times, the deviated points will gradually return to the real track trajectory
true turning ofAlthough
direction. the mobile theusers,
initial and the other
positioning errorfour points marked
is approximately 4.2with
m, theellipses are “pseudo-turning.”
positioning accuracy can
ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008 16 of 19
Therebe aregradually
seven “turning trends” on the routes in total, marked
compensated to some degree with the update of Wi-Fi positions. with a red circle in Figure 13.
However, the
divergence problem of point positions still exists at the corners.
Sum - Gyroscope Data
In Scheme 4, the noise matrix value Q is dynamically set according to the pedestrian movement
25

attributes (straight or turning). Figure 12 shows the accumulated gyro value corresponding to each
20
step of the pedestrian, in which the singular points in the three rectangular boxes represent the three
occasions of true turning of the mobile users, and the other four points marked with ellipses are
15

“pseudo-turning.” There are seven “turning trends” on the routes in total, marked with a red circle
in Figure 13. 10

-5

-10
0 50 100 150 200 250 300 350

Figure
Figure 12. 12. Analysisofofthe
Analysis the accumulated
accumulated gyro value.
gyro value.

6
x 10
3.7879

3.7879

3.7879
 -5

-10
0 50 100 150 200 250 300 350

ISPRS Int. J. Geo-Inf. 2016, 5, 8 Figure 12. Analysis of the accumulated gyro value. 16 of 19

6
x 10
3.7879

3.7879

3.7879

3.7879

3.7879

3.7879

3.7879

3.7879
5.133 5.1332 5.1334 5.1336 5.1338 5.134 5.1342
5
x 10

Figure
Figure 13. Analysis of
13. Analysis of the
the accumulated
accumulated gyro
gyro value.
value.

The experimental results show that the convergence speed at corners in Scheme 4 is higher than
that of
of the
theclassic
classicEKF.
EKF.WhileWhile maintaining
maintaining thethe
stability andand
stability the the
strong ability
strong to resist
ability Wi-FiWi-Fi
to resist gross gross
error
and
errorother advantages,
and other advantages,Scheme 4 also4 weakens
Scheme also weakensthe divergence at the
the divergence at corners while
the corners improving
while the
improving
positioning
the positioning accuracy.
accuracy.
The detailed error analysis of the four different schemes in Figure 14 and Table 3 shows that the
average positioning
positioning errorerrorofofAWEPDR
AWEPDRisisthe thesmallest
smallestandandaccounts
accountsforfor
56.2%, 37.5%,
56.2%, andand
37.5%, 85.9% of the
85.9% of
average errorerror
the average values of theof
values other
the three
othermodels. The maximum
three models. average positioning
The maximum error of AWEPDR
average positioning error of
AWEPDRfor
accounts accounts for 26.8%,
26.8%, 81.2%, and 81.2%,
66.4%, and the 66.4%, and thepositioning
minimum minimum errorpositioning
accounts error
for accounts for
55.8%, 4.2%,
ISPRS Int. J. Geo-Inf. 2016, 5, 8; doi:10.3390/ijgi5010008 17 of 19
55.8%,
and 4.2%,
78.4% of and 78.4%three
the other of themodels.
other three models.

Figure
Figure 14. Error analysis
14. Error analysis of
of different
different indoor
indoor positioning
positioning methods.
methods.

Table 3. Error analysis of different indoor positioning methods.

Wi-Fi PDR WEPDR AWEPDR

Minimum error/m 0.36 5.14 0.28 0.22
Average error/m 4.09 6.08 2.74 2.32
Maximum error/m 19.35 6.46 7.96 5.25

In conclusion, the indoor positioning result can be controlled within 2.5 m by adaptive noise
EKF filtering fusion of Wi-Fi system positioning results and PDR, which meet the demand of
pedestrian indoor navigation and positioning.
ISPRS Int. J. Geo-Inf. 2016, 5, 8 17 of 19

Table 3. Error analysis of different indoor positioning methods.

Wi-Fi PDR WEPDR AWEPDR

Minimum error/m 0.36 5.14 0.28 0.22
Average error/m 4.09 6.08 2.74 2.32
Maximum error/m 19.35 6.46 7.96 5.25

In conclusion, the indoor positioning result can be controlled within 2.5 m by adaptive noise EKF
filtering fusion of Wi-Fi system positioning results and PDR, which meet the demand of pedestrian
indoor navigation and positioning.

5. Conclusions
Clustering processing is performed on the fingerprint data of indoor positioning by the APC
algorithm to reduce the traversal data volume in positioning and improve the positioning accuracy.
Based on this, the calculation method of adaptively determining the filtering system dynamic noise
to conduct the fusion calculation based on pedestrian movement (straight or turning) is proposed
according to the respective advantages of Wi-Fi and PDR positioning systems. The experimental results
show that the fusion algorithm can effectively solve the instability problems such as Wi-Fi positioning
jumping and clustering, overcome the positioning error and cumulative error of PDR caused by the
initial position error, and thus improve the reliability and robustness of indoor positioning.

Acknowledgments: This work was partially supported by the Fundamental Research Funds for the Central
Universities under grant number 2015XKMS051, a project funded by the Priority Academic Program Development
of Jiangsu Higher Education Institutions, and the Cooperative Innovation Center of Jiangsu Province, and
Open Funds for Jiangsu Key Laboratory of Resources and Environment Information Engineering under grant
number JS201307.
Author Contributions: The corresponding author Xin Li proposed the research, organized the entire experimental
program, and drafted the manuscript. Chunyan Liu and Liwen Zhang were responsible for field data collection
and drafted part of the manuscript. Jian Wang and Zhengkui Li performed the data analysis and were involved in
the writing of the manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Mahfouz, M.R.; Kuhn, M.J.; To, G.; Fathy, A.E. Integration of UWB and wireless pressure mapping in surgical
navigation. IEEE Trans. Microw. Theory Tech. 2009, 57, 2550–2564. [CrossRef]
2. Gu, Y.; Lo, A.; Niemegeers, I. A survey of indoor positioning systems for wireless personal networks.
IEEE Commun. Surv. Tutor. 2009, 11, 13–32. [CrossRef]
3. Hightower, J.; Borriello, G. Location systems for ubiquitous computing. Computer 2001, 34, 57–66. [CrossRef]
4. Li, F.; Zhao, C.; Ding, G.; Gong, J.; Liu, C.; Zhao, F. A reliable and accurate indoor localization method using
phone inertial sensors. In Proceedings of the 2012 ACM Conference on Ubiquitous Computing, Pittsburgh,
PA, USA, 5–8 September 2012; pp. 421–430.
5. Pahlavan, K.; Li, X.; Makela, J.-P. Indoor geolocation science and technology. IEEE Commun. Mag. 2002, 40,
112–118. [CrossRef]
6. Jimenez, A.; Seco, F.; Prieto, C.; Guevara, J. A comparison of pedestrian dead-reckoning algorithms using
a low-cost MEMS IMU. In Proceedings of the 2009 IEEE International Symposium on Intelligent Signal
Processing, Budapest, Hungary, 26–28 August 2009; pp. 37–42.
7. Mautz, R. The challenges of indoor environments and specification on some alternative positioning systems.
In Proceedings of the 6th Workshop on Positioning, Navigation and Communication, Hannover, Germany,
19 March 2009; pp. 29–36.
8. Marano, S.; Gifford, W.M.; Wymeersch, H.; Win, M.Z. NLOS identification and mitigation for localization
based on uwb experimental data. IEEE J. Sel. Areas Commun. 2010, 28, 1026–1035. [CrossRef]
9. Foxlin, E. Pedestrian tracking with shoe-mounted inertial sensors. IEEE Comput. Graph. Appl. 2005, 25, 38–46.
[CrossRef] [PubMed]
ISPRS Int. J. Geo-Inf. 2016, 5, 8 18 of 19

10. Colomar, D.S.; Nilsson, J.; Handel, P. Smoothing for Zupt-aided INSS. In Proceedings of the 2012 International
Conference on Indoor Positioning and Indoor Navigation (IPIN), Sydney, NSW, Australia, 13–15 November
2012; pp. 1–5.
11. Jiménez, A.R.; Seco, F.; Prieto, J.C.; Guevara, J. Indoor pedestrian navigation using an INS/EKF framework
for yaw drift reduction and a foot-mounted IMU. In Proceedings of the 7th Workshop on Positioning
Navigation and Communication, Dresden, Germany, 8 June 2010; pp. 135–143.
12. Harle, R. A survey of indoor inertial positioning systems for pedestrians. IEEE Commun. Surv. Tutor. 2013,
15, 1281–1293. [CrossRef]
13. Ruiz, A.R.J.; Granja, F.S.; Prieto Honorato, J.C.; Rosas, J.I.G. Accurate pedestrian indoor navigation by
tightly coupling foot-mounted IMU and RFID measurements. IEEE Trans. Instrum. Meas. 2012, 61, 178–189.
[CrossRef]
14. Adams, D. Introduction to inertial navigation. J. Navig. 1956, 9, 249–259. [CrossRef]
15. Jiménez, A.R.; Zampella, F.; Seco, F. Improving inertial pedestrian dead-reckoning by detecting unmodified
switched-on lamps in buildings. Sensors 2014, 14, 731–769. [CrossRef] [PubMed]
16. Widyawan; Pirklb, G.; Munarettoc, D.; Fischerd, C.; Ane, C.; Lukowiczb, P.; Klepalf, M.; Timm-Gielg, A.;
Widmerh, J.; Pesch, D.; et al. Virtual lifeline: Multimodal sensor data fusion for robust navigation in unknown
environments. Pervasive Mob. Comput. 2012, 8, 388–401. [CrossRef]
17. Li, Y.; Zhang, Y.; Lan, H.; Zhang, P.; Niu, X.; El-Sheimy, N. WiFi-aided magnetic matching for indoor
navigation with consumer portable devices. Micromachines 2015, 6, 747–764. [CrossRef]
18. Masiero, A.; Guarnieri, A.; Pirotti, F.; Vettore, A. A particle filter for smartphone-based indoor pedestrian
navigation. Micromachines 2014, 5, 1012–1033. [CrossRef]
19. Ghinamo, G.; Corbi, C.; Francini, G.; Lepsoy, S.; Lovisolo, P.; Lingua, A.; Aicardi, I. The MPEG7 visual search
solution for image recognition based positioning using 3D models. In Proceedings of the 27th International
Technical Meeting of The Satellite Division of the Institute of Navigation, Tampa, FL, USA, 8–12 September 2014.
20. Saeedi, S.; Moussa, A.; El-Sheimy, N. Context-aware personal navigation using embedded sensor fusion in
smartphones. Sensors 2014, 14, 5742–5767. [CrossRef] [PubMed]
21. Ling, P.; Chen, R.; Liu, J. Using Inquiry-based bluetooth RSSI probability distributions for indoor positioning.
J. Glob. Position. Syst. 2010, 9, 122–130.
22. Youssef, M.A.; Agrawala, A.; Shankar, A.U.; Noh, S.H. A Probabilistic Clustering-Based Indoor Location
Determination System. Available online: http://www.cs.umd.edu/~moustafa/papers/locdet_tr.pdf
(accessed on 23 April 2015).
23. Yousseff, M.; Agrawala, A. Handling samples correlation in the Horus system. In Proceedings of the 23rd
Annual Joint Conference of the IEEE Computer and Communications Societies, Hong Kong, China, 7–11
March 2004.
24. Yousseff, M.; Agrawala, A. Location determination via clustering and probability distributions.
Pervasive Comput. Commun. 2003, 8, 143–150.
25. Shin, B.; Lee, J.H.; Lee, T.; Kim, H.S. Enhanced weighted K-nearest neighbor algorithm for indoor Wi-Fi
positioning systems. In Proceedings of the 8th International Conference on Computing Technology and
Information Management (ICCM), Seoul, South Korea, 24–26 April 2012.
26. Kuo, S.P. Cluster-enhances techniques for pattern-matching localization systems. Mob. Ad-Hoc Sens. Syst.
2007, 7, 1–9.
27. Bahl, P.; Padmanabhan, V.N. RADAR: An in-building RF-based user location and tracking system. In
Proceedings of the 9th Annual Joint Conference of the IEEE Computer and Communications Societies,
Tel Aviv, Israel, 26–30 March 2000; pp. 775–784.
28. Klingbeil, L.; Wark, T. A wireless sensor network for real-time indoor localization and motion monitoring. In
Proceedings of the 7th International Conference on Information Processing in Sensor Networks, St. Louis,
MO, USA, 22–24 April 2008; pp. 39–50.
29. Frey, B.J.; Dueck, D. Clustering by passing messages between data points. Science 2007, 315, 972–976.
[CrossRef] [PubMed]
30. Yu, F.; Jiang, M.; Liang, J. An indoor localization of WiFi based on branch-bound algorithm. In Proceedings
of the 2014 International Conference on Information Science, Electronics and Electrical Engineering (ISEEE),
Sapporo, Japan, 26–28 April 2014; pp. 1306–1308.
ISPRS Int. J. Geo-Inf. 2016, 5, 8 19 of 19

31. Song, L.; Hatzinakos, D. A cross-layer architecture of wireless sensor networks for target tracking.
IEEE/ACM Trans. Netw. 2007, 15, 145–158. [CrossRef]
32. Cho, S.Y.; Park, C.G. MEMS based pedestrian navigation system. J. Navig. 2006, 6, 135–153. [CrossRef]
33. Klepal, M.; Beauregard, S. A backtracking particle filter for fusing building plans with PDR displacement
estimates. In Proceedings of the 5th Workshop on Positioning, Navigation and Communication, Hannover,
Germany, 27 March 2008; pp. 207–212.
34. Chen, G.; Meng, X.; Wang, Y.; Zhang, Y.; Tian, P.; Yang, H. Integrated WiFi/PDR/Smartphone using an
unscented Kalman Filter algorithm for 3D indoor localization. Sensors 2015, 15, 24595–24614. [CrossRef]
[PubMed]
35. Chen, W.; Chen, R.; Chen, Y.; Kuusniemi, H.; Wang, J.; Fu, Z. An effective pedestrian dead reckoning
algorithm using a unified heading error model. In Proceedings of the 2010 IEEE/ION Position Location and
Navigation Symposium (PLANS), Indian Wells, CA, USA, 4–6 May 2010; pp. 340–347.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons by Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).