Hindawi Publishing Corporation

International Journal of Distributed Sensor Networks

Volume 2014, Article ID 371350, 22 pages
http://dx.doi.org/10.1155/2014/371350

Research Article
A Location Estimation Algorithm Based on
RSSI Vector Similarity Degree

Fengjun Shang,1 Wen Su,1 Qian Wang,2 Hongxia Gao,1 and Qiang Fu1
1
College of Computer Science and Technology, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
2
Geophysical Exploration Academy, China Metallurgical Geology Bureau, Baoding 071051, China

Correspondence should be addressed to Fengjun Shang; [email protected]

Received 22 April 2014; Revised 2 July 2014; Accepted 10 July 2014; Published 1 September 2014

Academic Editor: Lei Shu

Copyright © 2014 Fengjun Shang et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We present a detailed study on the RSS-based location techniques in wireless sensor networks (WSN). There are two aspects in this
paper. On the one hand, the accurate RSSI received from nodes is the premise of accurate location. Firstly, the distribution trend of
RSSI is analyzed in this experiment and determined the loss model of signal propagation by processing experimental data. Secondly,
in order to determine the distance between receiving nodes and sending nodes, Gaussian fitting is used to process specific RSSI
at different distance. Moreover, the piecewise linear interpolation is introduced to calculate the distance of any RSSI. On the other
hand, firstly, the RSSI vector similarity degree (R-VSD) is used to choose anchor nodes. Secondly, we designed a new localization
algorithm which is based on the quadrilateral location unit by using more accurate RSSI and range. Particularly, there are two
localization mechanisms in our study. In addition, the generalized inverse is introduced to solve the coordinates of nodes. At last,
location error of the new algorithm is about 17.6% by simulation experiment.

1. Introduction truly. Obviously, the location information of the node is the

premise of sensor nodes that are perceived and collected
Localization is certainly needed for robots to efficiently carry data. It is imprecise and lacking practical significance that
out given tasks such as cleaning, serving, and guiding. It is a there is only perception data of nodes and there is not the
challenging research topic in mobile robot which has received perception of location information which signed the source
much attention [1]. Moving around without location for of perception data [5]. Therefore, it is important to locate the
robots is the same as walking with closed eyes for people [2]. node in WSN, so the location technology also is becoming
They do not know where they are and which direction they an essential function and a critical supporting technology in
should move to. Under this background, reliable and efficient WSN [6]. In terms of the collection of network data and event
localization is still a critical issue for mobile robots. Research monitoring, the location information of nodes is important.
on localizing mobile robots has come to prominence in In the field of remote sensing, it is significant to track
wireless sensor network over the last years [3]. and monitor. It is contributed to compare composited route
To a large extent, WSN has changed the current situation and optimize communication task and improve composited
that people get information by the traditional way, such as routing table [7]. Some domestic scholars also research on
their sense of touch, sight, and smell. WSN has become a deep discussions of localization in WSN [8, 9]. Unknown
new formula of technology to get data. Wherever in the field nodes and anchor nodes are two important parts of location
of national security or national economy building, WSN has system. In addition, in terms of choosing anchor nodes,
been applied widely [4]. In a variety of applications in WSN, some researchers pay attention to research on RSSI-similarity
only if the location information of the node itself is combined degree [10].
with the information which is captured and collected from There are two existing formulas of methods of node local-
sensor nodes “the location of the incident” can be illustrated ization in WSN by measuring distance or angle information
accurately and the region of target monitoring is reflected between nodes in the location process and not measuring
2 International Journal of Distributed Sensor Networks

information: the method is based on range location (Range- coordinate of the unknown node [17, 18], which is formed by
based) and without range location (Range-free) [11]; the anchor nodes.
information of range or angle between communication nodes In [19], Approximate Point In triangulation Test (APIT) is
is obtained; it is the premise of the former method to locate mentioned. It is a concrete application of centroid localization
nodes; the latter method is not needed to measure range method; the location principle is as follows: there are any
information between nodes directly; it is estimated coordi- three anchor nodes that composed one triangle; there is
nates of nodes just by getting information of communication a collection formed by all triangles which included the
hops between nodes or network connectivity. There are unknown node; the coordinate of the unknown node is
common ranging methods as follows: TOA ranging method determined by determining the center of mass of graphic
[9, 11], TDOA ranging method [11], AOA ranging [12], and which is composed of all triangles of the collection.
RSSI ranging method [13]. Centroid algorithm is the most popular algorithm in
Range is the premise of location; precise range is the many location methods in WSN, because the easy operation
assurance of accurate location. Only a certain number of and the characteristics of few errors are included in the
reference nodes (anchor nodes) are combined with effective algorithm. Yedavalli and Krishnamachari has put forward
localization algorithm if they ensured feasible location task, sequence of location algorithm [20] (SBL, Sequence-based
lots of equipment is included in the process; they are arranged Localization); Liu et al. has put forward a new SBL algorithm;
before and are able to communicate. At the same time, in it combined SBL with three orthocenter algorithm [21]; it is a
the place of receiving nodes, information of signal strength concrete instance of centroid localization method; it achieved
and angle between nodes is needed to measure and convert good localization accuracy, but it is needed to improve on
it into distance information. With the help of geometric or boundary node location or ideal environment.
mathematical relationships we can carry out the location The location principle of convex programming location
task after achieving information of distance or orientation method [22] (convex optimization) is that the whole network
between nodes. In typical location algorithms, the method is regarded as a model of convex collection by using the
of Range-based mainly included trilateral location and trian- network connectivity; it is through the way of bound com-
gulation location and maximum likelihood location; Range- bination and plans to determine the possible region within
free mainly included centroid location, convex programming the unknown node and estimate coordinates.
location, and DV-Hop location method. In [23], researchers have analyzed location methods of
The principle of Hop count [12] ranging method is based DV-Hop and DV-distance; the experimental result has shown
on hop counts between anchor nodes and unknown node that the location accuracy can achieve 20% when the network
and the product of distance of mean hop in the network connectivity is 9 and the proportion of anchor nodes is 10%,
to determine their distance. Information of hop count is but the location accuracy is declining evidently along with
obtained as follows: anchor nodes send a packet which the increasing range error. Zhang and Wu have put forward
included lots of information such as its location information a modified location algorithm [24] which is based on the
to the network, if receiving nodes received the packet, it will estimation of mean hop distance and location correction; it
add one hop to itself and retransmission, until the unknown has solved the situation that the distance of mean hop cannot
node receives this packet; at this time, the minimum hop (the reflect the real distance, which is estimated by the single
shortest path) is regarded as the hop count between anchor anchor node in the network. The result of experiment has
nodes and the unknown node. This range method requires shown that mean error of the modified algorithm is reduced
the uniform distribution of nodes in the network; it can about 8.8497% and 14.457%; it has achieved better location
ensure that the distance of mean hop in the network better accuracy.
reflected the layout of nodes and achieved higher ranging In this paper, a novel location algorithm based on RSSI
precision. vector similarity degree is presented and a location system is
The coordinate of the unknown node is estimated by the designed, which is applied to locate sensor nodes indoor. Our
trilateral location method; it has got good effect [14]. proposed contributions are as follows.
In [15], the value of nodes localization is achieved by the
steepest descent algorithm which is a modified algorithm (1) The Gaussian fitting has optimized the value of RSSI
of maximum likelihood estimation method; moreover, the at different distance.
location accuracy and smaller computational cost can be
obtained from the steepest descent algorithm which has (2) The linear interpolation is used to decrease the calcu-
obvious effect. In [16], the maximum likelihood estimation lation of RSSI; it is according to the relation between
and the Kalman filtering are composited; it has a double different RSSI and distance.
effect on node prelocation and tracking, and it also has higher
(3) The vector similar degree is helpful in choosing
location accuracy.
anchor nodes.
The location principle of centroid localization is as fol-
lows: the unknown node sends broadcast messages to the net- (4) The generalized inverse is solved to estimate the
work for determining its location information; after anchor coordinate of unknown node by equations.
nodes received broadcast messages, they sent response of
information of its own location to the unknown node; the (5) The location system is able to work well indoor when
center of mass in this graphic is regarded as the estimation the mobile node is dynamic.
International Journal of Distributed Sensor Networks 3

2. Ranging Method of RSSI-Based path loss of receiving sign when the reference distance is 𝑑0 ;
𝑛 indicated the path loss index in a specific environment; it
Received signal strength indicator (RSSI) indicated the indicated the speed of the path loss, which is increased along
energy loss in the process of signal transmission; the RSSI with increasing distance; 𝑋𝜎 is in dB; it is a cover factor when
value is associated with the size of signal attenuation. In the the range of standard deviation 𝜎 is 4∼10 and the mean value
process of signal transmission, the smaller the RSSI value is 0; the larger the 𝜎, the greater the uncertainty of the model.
the less the attenuation. Usually, the RSSI ranging [25] is The signal strength of receiving nodes is as follows:
based on the experience model or theory model. The RSSI
ranging of model-based experience has shown that an offline RSSI = 𝑃𝑡 − 𝑃𝐿 (𝑑) . (3)
database is built between few RSSI values of nodes which
are known location information and their distance of signal In this formula, 𝑃𝑡 indicated the signal transmission power,
propagation; in the process of nodes location, the stored data 𝑃𝐿 (𝑑) indicated the path loss when the distance is 𝑑, and
in the database is constantly compared to implement nodes they are both in dBm. 𝐴 indicated the signal strength which
location. The RSSI ranging model-based theory implemented is received from reference nodes at the distance 𝑑0 ; 𝐴 is as
nodes localization by determining the environmental param- follows:
eters in the loss model of signal propagation and plugging
the RSSI value which is received from receiving nodes into 𝐴 = 𝑃𝑡 − 𝑃𝐿 (𝑑) . (4)
the model, so as to estimate distance between nodes. There
is a key of former method; lots of measuring work of off-line The path loss model is as follows, which is measured at the
in the region is required; in addition, the location accuracy real distance 𝑑 (m):
is limited by the number of nodes and measuring workload.
𝑑
The core of the latter method lied in researching on critical 𝑃 (𝑑) = 𝑃 (𝑑0 ) − 10𝑛 log ( ) − 𝑋𝜎 . (5)
parameters of the model detail by the accurate measuring 𝑑0
of RSSI value; hardware conditions and actual environment
In this formula, 𝑃(𝑑) indicated the received signal strength
can be better reflected; in addition, location accuracy and
when the measured real distance is 𝑑 (m). 𝑃(𝑑0 ) indicated
improved location performance can be enhanced.
received signal strength when the reference distance is 𝑑0 ,
Common methods of RSSI ranging are as follows, which
𝑋𝜎 ∼ 𝑁(0, 𝜎2 ).
are based on the theory model: the path loss model of free
We take the reference distance 𝑑0 = 1 m; it can be
space propagation [26] and the block model of logarithmic
obtained from formula (3) and formula (5).
normal [26] (Shadowing model), and so forth.
The path loss model of free space propagation is an ideal 𝑑
transmission case, it is known that there is an infinite vacuum RSSI = 𝐴 − 10𝑛 log ( ) − 𝑋𝜎 . (6)
𝑑0
around antenna, the signal transmission energy is only
related to transmission distance, there is a linear relationship RSSI is regarded as RSSI value of multiple times of measure-
between the signal transmission energy and transmission ment
distance, this model has no effect on obstacles and scattered
reflection, and so forth. The path loss model [26] is as follows: RSSI = 𝐴 − 10𝑛 log (𝑑) . (7)
Loss = 32.44 + 10𝑛 log (𝑑) + 10𝑛 log (𝑓) . (1) 𝑑 is regarded as the undetermined distance
In this formula, loss indicated the path loss of signal energy,
𝑑 = 10(𝐴−RSSI)/10𝑛 . (8)
𝑑 indicated the signal transmission distance (m), 𝑓 indicated
the wireless signal frequency (MHz), and 𝑛 indicated the path The Shadowing model is chosen to indicate the power atten-
attenuation factor in the actual environment. uation in the process of the wireless signal transmission; due
However, the application environment of wireless sensor to various obstacles and interference factors, there are kinds
signal is not in a free space, but in the actual environment of the wireless signal propagation environment; it can be
such as industrial sites or indoor buildings; it needs to reflected the real measurement environment by determining
consider shade and absorbance by obstacles and the inter- related environmental parameters in the model.
ference of scattered reflection, and so forth. The attenuation
characteristic of channels in the long distance is following
the lognormal distribution; it is commonly used by the block 2.1. RSSI Ranging Method Based on Gaussian Fitting. As
model of logarithmic normal; the path loss model is as shown in Shadowing model, there is a corresponding rela-
follows: tionship between wireless signal propagation loss and trans-
mission distance in the actual test environment. It means
𝑑 that according to the received information of signal strength
𝑃𝐿 (𝑑) = 𝑃𝐿 (𝑑0 ) + 10𝑛 log ( ) + 𝑋𝜎 . (2) the distance between sending nodes and receiving nodes can
𝑑0
be obtained. However, the conclusion of experiment cannot
In this formula, 𝑃𝐿 (𝑑) indicated the path loss of receiving rely on one time of measurement task, but on the basis of
sign when the measuring distance is 𝑑 (m), it indicated the a large number of test data. So, in terms of the same node
absolute power value, and it is in dBm; 𝑃𝐿 (𝑑0 ) indicated the and the same distance, received signal strengths are taking
4 International Journal of Distributed Sensor Networks

many times of measures. On the basis of measurement, the 75

distribution trend of RSSI in the test environment is analyzed
and interference data and error data are eliminated; the most
70
representative signal strength value at specific distance is
obtained; signal strength is regarded as the basis of distance

RSSI (dBm)
estimation at this place. 65

2.1.1. Experimental Environment. The location system is 60

included the hardware platform and software platform. Hard-
ware platform included anchor nodes and unknown node, 55
gateway control equipment. The measurement is worked on
the perception platform of SensorRF107H2 0, the sensing 50
platform is designed by the wireless gateway which owned
the core of high-performance microcontroller and in 32-bit
0 50 100 150 200 250 300 350 400 450
ARM, and it included advanced designing tools of software
Measurement times
and hardware such as online emulators when users develop,
debug, and test software. Anchor nodes and unknown node d = 5 m RSSI data
are composed of modules of CC2530 of the system of ZigBee
which is included TI. Figure 1: RSSI experimental data of node at 𝑑 = 5 m.

2.1.2. Acquisition of RSSI Data. Signal strength indicator

(RSSI) is read from the register RSSI VAI in the data packets 70
which are received from sensor receiving nodes, a sensor
node is set and regarded as a receiving device at intervals
65
of 1 m, RSSI data is measured more than 400 times at every
place, and each sensor node is received and recorded its signal
RSSI (dBm)

strength value. Receiving nodes only can receive few signal 60

strength values beyond 25 m; parts of signal value are not
detected, because the RSSI value of characterization signal
attenuation is included in data packets, so it is meaningless 55
to keep measuring. In this paper, the range of RSSI ranging
is kept 0∼25 m. For example, the distance between sending
node and receiving node is taken 5 m and 10 m, respectively, 50
we have made statistics for collecting RSSI value and chosen
more than 400 times of measurement to be a sample and got
0 100 200 300 400 500 600
the characteristics picture of RSSI, and they are as in Figures
Measurement times
1 and 2.
Figures 1 and 2 can be seen, with the passage of time;
d = 10 m RSSI data
although RSSI generally presented obvious nonstationary
characteristics, the RSSI value in certain distance are always Figure 2: RSSI experimental data of node at 𝑑 = 10 m.
fluctuated up and down between the constant real RSSI value,
the closer distance from emission node, and the frequenter
change of measurement value; on the contrary, the farther the
distance from emission node, the gentler the change. Environment factor 𝑛 indicated, along with the increasing
distance, the speed of signal loss in the process of actual trans-
2.1.3. Shadowing Model. In order to exactly describe the mission. They are satisfied with the following relationship:
actual measurement environment and ensure the accuracy
of RSSI ranging, the parameters of signal propagation model
𝐴 − RSSI
are needed to define. 𝐴 indicated the received signal strength 𝑛= , (9)
value when the reference distance is 𝑑0 = 1 m and 𝑛 indicated 10 ∗ log (𝑑)
the path loss index.
The determination of parameter 𝐴 is as follows. where 𝑛 is obtained by generating into more than 3000 groups
𝐴 = 𝑃(𝑑0 ), the radio frequency parameter 𝐴 indicating of data in the range of 2 ∼ 10 m; it is as in Table 1.
the received absolute value of mean energy, which is received The mean value 𝑛 can be obtained as follows:
from the projection node when distance is 1 m; it is in dBm.
When the distance is 1 m, 500 groups of data are continuously 10
measured; finally, there is a result that 𝐴 = 117. 𝑛 = ∑ 𝑛𝑖 = 6.9861. (10)
The determination of parameter 𝑛 is as follows. 𝑖=2
International Journal of Distributed Sensor Networks 5

Table 1: The value of 𝑛.

Distance
2m 3m 4m 5m 6m 7m 8m 9m 10 m
𝑛 5.982 5.904 5.421 7.908 6.670 10.820 6.496 7.507 6.070

1 1
0.9 0.9
0.8 0.8
0.7 0.7
Probability density

Probability density
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
−2 0 2 4 6 8 −2 0 2 4 6 8 10
d (m) d (m)

RSSI = 117 pdf RSSI = 89 pdf

Figure 3: First group of RSSI probability density. Figure 4: Second group of RSSI probability density.

1
The Shadowing transmission model can be obtained by
combining with parameter 𝐴 and parameter 𝑛; it is as follows: 0.9
0.8
𝑑
𝑃 (𝑑) = 117 − 10 ∗ 6.9861 ∗ log ( ) − 𝑋𝜎 . (11)
𝑑0 0.7
Probability density

0.6
2.1.4. The Distribution Trend of RSSI and Gaussian Fitting. 0.5
In order to analyze the distribution trend of RSSI, many
groups of RSSI data are integrated and analyzed and drew 0.4
four groups of probability density curves when RSSI is, 0.3
respectively, 117 DBM, 89 DBM, 61 DBM, and 25 DBM; the
0.2
results are as in Figures 3, 4, 5, and 6.
It can be seen from Figures 3, 4, 5, and 6 of RSSI 0.1
probability density that the distribution of RSSI value of the 0
real measurement presented a probability distribution; there 0 2 4 6 8 10
are some characteristics are as follows. d (m)

RSSI = 61 pdf
Concentration. The peak of the curve (the mean location) is
located in the central. Figure 5: Third group of RSSI probability density.

Basic Symmetry. Two sides of curves are basic symmetry and

ends of the curve are closed to horizontal axis. The centre of
curves is based on mean value. probability density of RSSI and normal distribution [30].
So Gaussian fitting is used to fit RSSI value in experiment
Volatility. The trends of two sides of curves are declined and probability error and invalid data are eliminated, the
gradually at the place of mean value. peak value of probability is found, the estimating distance is
Some scholars also research on the kind of distribution extracted, which is closer to the real distance, and the process
of RSSI [27–29]; they think RSSI value approximately fol- is prived an experiment basis to get accurate location.
lowed normal distributions as a whole. According to the If RSSI data is closed to normal distribution, the peak
experimental data and probability density in this paper, it can of normal distribution will be the place of the greatest
be seen that there is a similarity between actual measured probability density, and the RSSI value is most likely the
6 International Journal of Distributed Sensor Networks

1 If all errors are following normal distribution, the probability

0.9 𝑃(𝑒1 , 𝑒2 , . . . , 𝑒𝑚 ) will be changed; there is a formula as follows:
0.8
1 𝑚 (𝑓 − 𝑓 ̂ )2
𝑖 𝑖
𝑃 (𝑒1 , 𝑒2 , . . . , 𝑒𝑚 ) = exp [− ∑ 2
] . (16)
0.7 𝜎√2𝜋 2𝜎
Probability density

𝑖=1
[ ]
0.6
The fitting curve will be the best fitting form of the original
0.5
curve, if 𝑃(𝑒1 , 𝑒2 , . . . , 𝑒𝑚 ) got the peak of normal distribution.
0.4 At the same time, the sum of squares of fitting residual is
0.3
minimal; there is a formula as follows:
𝑚 𝑚
2
0.2 𝑆 = ∑ (𝑓𝑖 − 𝑓̂𝑖 ) = ∑ 𝑒𝑖 2 . (17)
0.1 𝑖=1 𝑖=1

0 Formula (18) can be described as

−2 0 2 4 6 8 10
𝑚
d (m) 2
𝑆 = ∑ [𝑓𝑖 − 𝑔 (RSSI𝑖 , 𝑎̂1 , 𝑎̂2 , . . . , 𝑎̂𝑁)] . (18)
RSSI = 25 pdf 𝑖=1

Obviously, there are relationships as follows:

Figure 6: Fourth group of RSSI probability density.
𝜕𝑆 𝜕𝑆 𝜕𝑆
= 0, = 0, . . . , = 0,
𝜕𝑎1 𝜕𝑎2 𝜕𝑎𝑁
corresponding distance. So the Gaussian fitting [31] is used to 𝑚
𝜕𝑓
fit the original RSSI data by looking for the peak of probability ∑ [𝑓𝑖 − 𝑔 (RSSI𝑖 , 𝑎̂1 , 𝑎̂2 , . . . , 𝑎̂𝑁)] ( ) = 0,
density; it is aimed at reducing the interference that small 𝑖=1 𝜕𝑎1
probability events have an effect on the overall measurement 𝑚
process. In order to improve the precision of RSSI ranging, 𝜕𝑓
∑ [𝑓𝑖 − 𝑔 (RSSI𝑖 , 𝑎̂1 , 𝑎̂2 , . . . , 𝑎̂𝑁)] ( ) = 0, (19)
the least square [32] is used to fit curves of probability density 𝑖=1 𝜕𝑎2
of RSSI.
(RSSI𝑖 , 𝑓𝑖 ) is described in a two-dimensional coordinate ..
.
system; RSSI𝑖 indicated the 𝑖th value of signal strength; 𝑓𝑖
indicated the peak value of probability density, which is 𝑚
𝜕𝑓
correspondent to RSSI𝑖 ; there is a relationship as follows: ∑ [𝑓𝑖 − 𝑔 (RSSI𝑖 , 𝑎̂1 , 𝑎̂2 , . . . , 𝑎̂𝑁)] ( ) = 0,
𝑖=1 𝜕𝑎𝑁
𝑓 = 𝑔 (RSSI, 𝑎1 , 𝑎2 , . . . , 𝑎𝑁) . (12) where 𝑎1 , 𝑎2 , . . . , 𝑎𝑁 is determined by solving the above
equation; at the same time, the least square solutions of the
In this formula, 𝑎1 , 𝑎2 , . . . , 𝑎𝑁 indicated 𝑁 pending constant. equation is obtained.
There is an assumed fitting function between RSSI𝑖 and 𝑓𝑖 Figure 7 has shown the probability density of fitting when
as follows: RSSI = 117 DBM. Since then, the RSSI value at the highest
peak of curve is regarded as the signal strength value (distance
𝑓̂ = 𝑔 (RSSI, 𝑎̂1 , 𝑎̂2 , . . . , 𝑎̂𝑁) . (13) 1 m). Similarly, in terms of any RSSI value, the distance
is regarded as the basis of distance estimation, which is
In this formula, 𝑎̂1 , 𝑎̂2 , . . . , 𝑎̂𝑁 indicated the optimal estimat- correspondent to the peak of probability density.
ing value which is correspondent to constants of original The Gaussian fitting function protoformula is as follows:
function. If fitting points are falling on the curve of the −2(𝑥 − 𝜇)
2
original function, there will be a relationship as follows: 𝑓 (𝑥) = 𝑎 ∗ 𝑒 ,
𝜔2
(20)
𝑓𝑖 = 𝑔 (RSSI𝑖 , 𝑎1 , 𝑎2 , . . . , 𝑎𝑁) (𝑖 = 1, 2, 3, . . . , 𝑚) . (14) 2
∑𝑛𝑖=1 RSSI𝑖 𝑛
√ ∑𝑖=1 (RSSI𝑖 − 𝜇)
𝜇= , 𝜔= .
However, under normal circumstances, the original function 𝑛 𝑛−1
and the fitting function do not have to be same, there is a The fitting probability density curve of RSSI = 117 DBM is as
residual relationship as follows: follows:
2
𝑓 (𝑥) = 0.4643 ∗ 𝑒((𝑥−117)/1.214) . (21)
𝑒𝑖 = 𝑓𝑖 − 𝑓̂𝑖 (𝑖 = 1, 2.3, . . . , 𝑚) . (15)
Figure 7 has shown that the maximum value of the original
The smaller the residual 𝑒𝑖 the smaller the difference between probability density is RSSI = 117 dBm, but the maximum
fitting function and original function; 𝑓𝑖 can be better probability value of fitting is 𝜇 = 117 dBm. There is the same
reflecting the distribution trend of RSSI𝑖 at different distance. effect between fitting and original.
International Journal of Distributed Sensor Networks 7

0.5
120
0.45
0.4 100
0.35

RSSI (dBm)
80
0.3
0.25 60
0.2
0.15 40

0.1
20
0.05
0
114 115 116 117 118 119 120 121 122 123 0 5 10 15 20 25
Figure 7: The gauss curve fitting when RSSI = 117. d (m)

Figure 8: The trend of curves of RSSI value between experiment and

fitting.
2.2. Ranging Method Based on Interpolation. 400 groups of
RSSI data are taken at intervals of permanent distance; in
Figure 8, the blue line is the real measurement value of
RSSI in different distance; it can be seen that, with the
increasing distance, different distance of RSSI overall has 0.4
shown the trend of exponential attenuation, but there is a
Probability density

certain fluctuation which is influenced on interference factors 0.3

around. The green line is the fitting value of RSSI in different
distance. 0.2
Due to the fact that the trend of RSSI value is in accord
0.1
with the distribution characteristics of indexes, the least
squares are also used to fit the indexes function between the 0
relation of RSSI-d, 0.0001237 and 128.2 are presented in this 30
formula when the range of RSSI value is taken from 117 DBM 20 120
10 100
to RSSI = 61 DBM, and there is a formula of fitting curve as d (m 0
) 80
follows: Bm)
−10 60 I (d
RSS
𝑦 (𝑥) = 0.0001237𝑒0.4185∗𝑥 + 128.2𝑒−0.1224∗𝑥 . (22)
Figure 9: The curve of probability density when RSSI = 117 and RSSI
Figure 8 has shown the relationship between the experimen- = 61 DBM.
tal RSSI-d value and the fitting curve, respectively, the blue
line indicated RSSI curve and the green line indicated fitting
curve; in the graph, it can be seen that there is a deviation RSSI value is measured at intervals of 1 m and the distribution
between fitting result and experimental RSSI-d at individual of probability density is analyzed. The probability density
distance, but on the whole, it tended the exponential decay. curves when RSSI = 117 DBM and RSSI = 61 DBM are drawn,
Due to the fact that the obvious transmission signal is quickly respectively, in the diagram. But it is uncertain that the not-
attenuated at close distance, there is a slow attenuation and measured RSSI value is at the distance of 1.5 m and 2.5 m;
increasing error when the distance is farther. From the similarly, in terms of a given RSSI value, the distance between
experimental figure it can be also seen that there is a better sending nodes and receiving nodes is uncertain. To this end,
effect when the fitting curve is within 6 m and the trend of the interpolation mode is built according to the measured
fitting error is smaller; along with the increasing distance, probability distribution of RSSI value; the distance between
the fitting effect has shown a certain deviation and the fitting nodes is estimated according to the received RSSI value.
error is increased. So there is a linear relationship of RSSI The Gaussian fitting is used to fit the probability distribu-
value between adjacent distances which is less than 10 m, tion of RSSI. There is a fitting function as follows:
the not-measured RSSI value can be obtained by using the
interpolation [33]. 2
It is known that the real RSSI value is close to Gaussian −2(RSSI𝑖 − RSSI)
𝑓 (RSSI𝑖 ) = 𝑓0 + 𝑎 ∗ 𝑒 ,
distribution at different distance by analyzing the above 𝜎2
section; the place of larger distribution density of RSSI is (23)
2
most likely to be the distance between receiving nodes and ∑𝑛𝑖=1 RSSI𝑖 𝑛
√ ∑𝑖=1 (RSSI𝑖 − 𝜇)
projection nodes. As shown in Figure 9, the corresponding RSSI = , 𝜎= .
𝑛 𝑛−1
8 International Journal of Distributed Sensor Networks

Table 2: The contrast of range error between RSSI-d interpolation

(f2 , RSSI2) model and Shadowing model.
f2
RSSI (dBm) range [117 61] [117 56]
Distant range (m) [1 5] [1 10]
(fi , RSSIi ) RSSI-d interpolation model 12.11% 36.09%
fi Shadowing model 15.69% 46.27%

f1
(f1, RSSI1)
according to (26), the distance 𝑑𝑖 which is most likely
correspondent to RSSI can be obtained.
In this experiment, in the range of 0∼25 m, all mean
values of signal strength and standard deviation are indicated
as follows:
RSSI
RSSI1 RSSIi 25
RSSI2 RSSI = ∑ RSSI𝑖 = 64.8864 dBm,
𝑖=1
Figure 10: The linear interpolation.
𝑛 2
∑𝑖=1 (RSSI𝑖 − RSSI)
𝜎=√ = 32.4792,
Parameters 𝑓0 and 𝑎 can be determined by two groups of 𝑛−1
specific RSSI value, such as RSSI = 117 DBM and RSSI =
61 DBM. The peak of probability density of RSSI is searched 𝑓(RSSI = 117)max = 0.3989, 𝑓(RSSI = 61)max = 0.1816,
at each distance; the distance is regarded as the estimation (27)
distance between receiving nodes and sending nodes, which
is corresponded to the peak. Along with the increasing and indicated the probability density when RSSI = 117 and
distance between nodes, the RSSI value between nodes is RSSI = 61 by generating the related data into formula (23), and
decreased, so there is a linear relation between the peak of the interpolation model of RSSI-d is built as follows. When
probability density and distance as follows: the distance is 1 m and RSSI is 117 DBM, 𝑎 = −0.3534 and
𝑏 = 0.401 are obtained from formula (25), and the equation
𝑑 = 𝑓(RSSI𝑖 )max ∗ 𝑘 + 𝑏. (24) is as follows:
Along with the changed distance, parameters 𝑘 and 𝑏 indi- 2

cated the changed tendency of the probability density; they 𝑑 = 0.401 + (−0.3534) ∗ 𝑒−2(RSSI𝑖 −64.8864) /1054.9 . (28)
can be obtained by generating data. The model of RSSI-d is
In terms of different RSSI-d models at different RSSI, they can
shown as follows [34]:
be used in interpolation. In order to extrapolate the RSSI-
2 d model to greater RSSI range, it is significant to estimate
−2(RSSI𝑖 − RSSI)
𝑑 = max (𝑓0 + 𝑎 ∗ 𝑒 ) ∗ 𝑘 + 𝑏. (25) the long distant, so we have extrapolated the range of RSSI
𝜎2 from 117 to 56 (corresponding distance is 1∼10 m). Finally, the
contrast between mean range error and mean ranging error
As shown in Figure 10, 𝑓1 max and 𝑓2 max indicated functions of Shadowing model is as in Table 2.
of RSSI1 and RSSI2 , respectively, so, in terms of for all From Table 2, it can be seen that Shadowing model
RSSI𝑖 ∈ (RSSI1 , RSSI2 ), if the corresponding equation 𝑓𝑖 max ∈ and RSSI-d interpolation model can obtain corresponding
(𝑓1 max , 𝑓2 max ) is needed to be obtained, there is a relationship distance by speculating the RSSI value; they achieved the
as follows: ranging effect. But from the ranging error, it can be known
𝑓𝑖 max − 𝑓1 max 𝑓 − 𝑓𝑖 max that mean range errors of RSSI-d interpolation model are less
𝛽= = 2 max . (26) than Shadowing model; it is benefit from linear interpolation
RSSI𝑖 − RSSI1 RSSI2 − RSSI𝑖
of short distance.
In formula (26), 𝛽 indicated the interpolation coefficient, in
terms of any couples of RSSI according to the linear rule; 3. The Location Algorithm Based on R-VSD
their coefficient 𝛽 is the same value. So, there is an equation
between different couples of RSSI. If 𝛽 > 1, it is indicated Along with growing demands of applications, there are
that the RSSI𝑖 is in the range of RSSI1 and RSSI2 , so the many kinds of methods, but every method has its merits.
extrapolation can be used; on the contrary, if 𝛽 < 1, the At present, the research on indoor location technology in
interpolation can be used. wireless is relatively concentrated on signal-based RF [35],
If it is known that RSSI𝑖 , the corresponding normal peak there are various technologies in wireless network, such
𝑓𝑖 max , is calculated by 𝑓𝑖 max = (1 − 𝛽) ∗ 𝑓1 max + 𝛽 ∗ 𝑓2 max ; as ultra wideband (UWB), Wi-Fi (IEEE 802.11), Bluetooth,
International Journal of Distributed Sensor Networks 9

and radio frequency identification (RFID). If researchers There is a collection by ordering RSSI value in the vector table
consider hardware conditions of location and signal resources as follows:
and location accuracy, the method based on the received
𝑅1󸀠
signal strength indicator (RSSI) [21] is widely used in all [ 󸀠]
of location methods. Existing localization algorithms [23] [ 𝑅2 ]
[ ]
can also obtain good location effect, but the researching is [ . ]
[ . ]
not completed on the part of edge nodes; in addition, there 𝑅󸀠 = [
[ .󸀠 ]
]
is a problem of large location errors which are caused by [ 𝑅𝑘 ]
[ ]
the pending location nodes that are outside of the unit. On [ .. ]
[ . ]
account of the above problems, an improved localization 󸀠
algorithm RSSI-based with vector similarity is proposed. In [𝑅𝑚 ]
the process of location, the method of R-VSD is used to 𝑋 (𝑅11 ) 𝑋 (𝑅12 ) ⋅ ⋅ ⋅ 𝑋 (𝑅1𝑗 ) ⋅ ⋅ ⋅ 𝑋 (𝑅1𝑛 )
choose optimal anchor nodes; the new location algorithm is [ ]
[ 𝑋 (𝑅21 ) 𝑋 (𝑅22 ) ⋅ ⋅ ⋅ 𝑋 (𝑅2𝑗 ) ⋅ ⋅ ⋅ 𝑋 (𝑅2𝑛 ) ]
used to estimate coordinates of unknown nodes. [ ]
[ .. .. .. .. ]
[ . . ⋅⋅⋅ . ⋅⋅⋅ . ]
[
=[ ].
[ 𝑋 (𝑅𝑘1 ) 𝑋 (𝑅𝑘2 ) ⋅ ⋅ ⋅ 𝑋 (𝑅𝑘𝑗 ) ⋅ ⋅ ⋅ 𝑋 (𝑅𝑘𝑛 ) ]
]
3.1. The RSSI Vector Similarity Degree. In order to describe [ ]
[ .. .. .. .. ]
the similar degree between the RSSI vector of unknown nodes [ . . ⋅⋅⋅ . ⋅⋅⋅ . ]
and the RSSI vector of reference sample points, the new 𝑋 (𝑅
[ 𝑚1 ) 𝑋 (𝑅𝑚2 ) ⋅ ⋅ ⋅ 𝑋 (𝑅2𝑗 ) ⋅ ⋅ ⋅ 𝑋 (𝑅𝑚𝑛 )]
indicator, similar degrees of RSSI vector is built; reference
(32)
sample points which are nearest to the unknown node
accurately can be found. 𝑋(𝑅𝑘𝑗 ) indicated the RSSI values of the 𝑗th ordering; it
is received from the 𝑘th reference sample points; namely,
Definition 1. If a node can receive radio signal from 𝑛 anchor anchor node which is the 𝑗th distance from the unknown
nodes, the received RSSI value can set a vector collection as node.
follows:
Definition 2. There are two different vectors 𝑅𝑇󸀠 =
{𝑋(𝑅𝑇1 ), 𝑋(𝑅𝑇2 ), . . . , 𝑋(𝑅𝑇𝑗 ), . . . , 𝑋(𝑅𝑇𝑛 )} and 𝑅𝑘󸀠 =
Ψ = {RSSI1 , RSSI2 , . . . , RSSI𝑖 } (𝑖 = 1, 2, . . . , 𝑛) . (29) {𝑋(𝑅𝑘1 ), 𝑋(𝑅𝑘2 ), . . . , 𝑋(𝑅𝑘𝑞 ), . . . , 𝑋(𝑅𝑘𝑛 )} which are formed
by independent keywords collection Ψ󸀠 = {𝑋(𝑅1 ), 𝑋(𝑅2 ), . . .
𝑋(𝑅𝑗 )}, so the deviation degree of keyword 𝑋(𝑅𝑇𝑝 ) between
In this formula, RSSI𝑖 indicated RSSI values which is received two vectors is as follows:
from 𝑖th anchor node node.
󵄨 󵄨
𝑑 (𝑝) = 󵄨󵄨󵄨󵄨𝑋 (𝑅𝑇𝑝 ) − 𝑋 (𝑅𝑘𝑞 )󵄨󵄨󵄨󵄨 , if 𝑅𝑇𝑝 = 𝑅𝑘𝑞 . (33)
RSSI value of the vector collection Ψ is ordered from big
to small; the collection Ψ is as follows: Definition 3. There are two RSSI vectors.
𝑅𝑇 = {RSSI𝑇1 , RSSI𝑇2 , . . . , RSSI𝑇𝑗 , . . . , RSSI𝑇𝑛 } and 𝑅𝑘 =
{RSSI𝑘1 , RSSI𝑘2 , . . . , RSSI𝑘𝑞 , . . . , RSSI𝑘𝑛 }; the normalization
Ψ󸀠 = {𝑋 (𝑅1 ) , 𝑋 (𝑅2 ) , . . . , 𝑋 (𝑅𝑗 )} (𝑗 = 1, 2, . . . , 𝑛) . processing is used to the RSSI vector 𝑅𝑇 :
(30) 𝑛
∑ RSSI𝑇𝑝 = 1, (𝑝 = 1, 2, . . . , 𝑛) . (34)
𝑗=1
The collection Ψ󸀠 indicated the key collection of RSSI value;
So, the similar degree 𝜌(𝑘) between the RSSI vector 𝑅𝑇 and
𝑋(𝑅𝑗 ) indicated the received RSSI value which is ranked the 𝑅𝑘 is as follows:
𝑗th; namely, anchor node is regarded as the key of RSSI value,
which is the 𝑗th distance from the unknown node. 𝑛
𝜌 (𝑘) = 𝜌 (𝑅𝑇 , 𝑅𝑘 ) = ∑ 𝑑 (𝑝) ∗ RSSI𝑇𝑝 ,
The RSSI vector which is formed by 𝑛 anchor nodes and 𝑝=1 (35)
𝑚 reference sample points is as follows:
(𝑘 = 1, 2, . . . , 𝑚) (𝑝 = 1, 2, . . . , 𝑛) 𝑅𝑘 ∈ 𝑅.

𝑅1 𝑅11 𝑅12 ⋅ ⋅ ⋅ 𝑅1𝑖 ⋅ ⋅ ⋅ 𝑅1𝑛 In this formula, 𝑋(𝑅𝑇𝑝 ) indicated the deviation degree of
[ 𝑅2 ] [ 𝑅21 𝑅22 ⋅ ⋅ ⋅ 𝑅2𝑖 ⋅ ⋅ ⋅ 𝑅2𝑛 ] keyword 𝑋(𝑅𝑇𝑝 ); RSSI𝑇𝑝 is the importance of RSSI at the
[ ] [ ]
[ .. ] [ .. .. . . ] place of 𝑇𝑝.
[ . ] [ . . ⋅ ⋅ ⋅ .. ⋅ ⋅ ⋅ .. ]
𝑅=[ ] [
[ 𝑅𝑘 ] = [ 𝑅𝑘1
] (𝑖 ∈ 𝑁) .
[ ] [ 𝑅𝑘2 ⋅ ⋅ ⋅ 𝑅𝑘𝑖 ⋅ ⋅ ⋅ 𝑅𝑘𝑛 ]
] According to Definition 3, the vector similarity is satisfied
[ .. ] [ .. .. . .. ] with the following relations:
[ . ] [ . . ⋅ ⋅ ⋅ .. ⋅⋅⋅ . ]
[𝑅𝑚 ] [𝑅𝑚1 𝑅𝑚2 ⋅ ⋅ ⋅ 𝑅𝑚𝑖 ⋅ ⋅ ⋅ 𝑅𝑚𝑛 ] (1) 𝜌(𝑅𝑇 , 𝑅𝑘 ) ≥ 0; it is established only when the
(31) deviation 𝑋(𝑅𝑇𝑝 ) (𝑝 = 1, 2, . . . , 𝑛) is 0;
10 International Journal of Distributed Sensor Networks

(2) the smaller the 𝜌(𝑅𝑇 , 𝑅𝑘 ), the smaller the difference

Start
between 𝑅𝑇 and 𝑅𝑘 ; there is a higher similar degree; End
on the other hand, the lower similar degree is shown.

𝜌(𝑘) (𝑘 = 1, 2, . . . , 𝑚) is ordered from small to big; then, Distributed N

Calculated
anchor nodes Un
in vector table 𝑅, 𝑁 vectors which are the most similar to unknown nodes estimation
the vector 𝑅𝑇 = {RSSI𝑇1 , RSSI𝑇2 , . . . , RSSI𝑇𝑗 , . . . , RSSI𝑇𝑛 } are coordinates of
unknown nodes
satisfied with the relation; it is as follows:
Created RSSI by
𝜌 (𝑘) (𝑘 = 1, 2, . . . 𝑁, 𝑁 < 𝑚) < min (𝜌 (𝑘) (𝑘 > 𝑁)) . descending sort Yes
(36)

In this paper, according to above definition of similar degree, No

i < Un?
Arranged RSSI by
𝑁 reference sample points are searched, which are the most descending sort
similar to the unknown node; the region of the known
node is narrowed constantly. After iterative times, the area
is regarded as the estimation regional of the unknown node,
which is formed by 𝑁 reference sample points which are the
similar degree 𝜌(𝑘), in order to describe the similar degree Unknown node Pi is
inside of the
No Implement
externA
between the unknown node and other reference sample positioning unit
points accurately and estimate the coordinate of the unknown
node accurately. According to the similar degree, these Yes
reference sample points are weighted. The mass center of this
region is regarded as the ultimate estimation coordinate of the Implemented
unknown node. The formula is as follows: internA

1 𝑁 1 𝑁 Figure 11: The algorithm flow chart.

(𝑥, 𝑦) = ( ∑ 𝑥𝑘 ∗ RSSI𝑇𝑘 , ∑ 𝑦𝑘 ∗ RSSI𝑇𝑘 ) . (37)
𝑁 𝑘=1 𝑁 𝑘=1

The method of node location based on RSSI included two

location mechanisms which are internal and external, the In conclusion, the algorithm complexity is significantly lower
coordinate of the unknown node is estimated by combining than time complexity 𝑂(𝑛6 ) of SBL.
with the similar degree of RSSI which is used to find reference In this paper, in terms of the triangle location unit 𝑆,
sample points which are the closest to the unknown node. The which has shown the initial area, three median lines are
algorithm flow chart is as in Figure 11. divided into four small triangles for the original triangle;
In Figure 11, firstly, there are a number of Un unknown there is a triangle called a median triangle, which is formed
nodes and the number of 𝑛 anchor nodes is distributed in by three median lines; the areas of four triangle are 𝑆/4. So,
the region of pending. Secondly, unknown nodes received according to the dividing way, the location area is narrowed
RSSI which are sent and created the RSSI vector table which to 𝑆/4 when the iterative algorithm is used; at the same time,
is arranged by descending sort. Thirdly, the top four anchor three reference sample points are added in this process. The
nodes are taken to determine the quadrilateral location area within the unknown node is restricted to 𝑆/256 when
unit. Fourthly, the unknown node 𝑝 is judged whether it it is divided to 4 times; at this time, reference sample points
is inside of the quadrilateral location unit or not. If it is the number of 16 are required. So, the region area within the
inside of the quadrilateral location unit, the internal location node will be compressed to 𝑆/4𝑘 when the times of repetitions
arithmetic is implemented; on the contrary, the external of narrowing the area is 𝐾, clearly, this kind of iteration is
location arithmetic is implemented. Finally, the number of convergence.
unknown nodes in the range of Un should be judged before
estimation coordinates of unknown nodes are calculated.
Otherwise, the unknown node had to judge that whether it 3.2. The Location Algorithm Based on the Quadrilateral Loca-
is inside of the quadrilateral location unit or not. tion Unit. There are four anchor nodes regarded as reference
The time consuming of points collection and vector table anchor nodes, which are the closest to the unknown node.
are both 𝑂(𝑛) (𝑛 is the number of anchor nodes). The time The quadrilateral is regarded as the location region, which
of arranging consumed is 𝑂(𝑛2 ). There are three times of is formed by these four anchor nodes. The unknown node is
iterative loop when the area is narrowed and when it is judged determined whether inside of the quadrilateral or outside of
whether it is inside of the quadrilateral location unit or not. the quadrilateral by the relation of area constraint. Under this
In the process of location arithmetic, it is required to compare section, these problems are solved: how to select reference
the RSSI vector of the unknown node with the respective RSSI sample point if it is inside of quadrilateral; if it is outside of the
vector of six reference sample points when each unknown quadrangle, how to determine the coordinate of the unknown
node is needed to be located; the time consuming is 𝑂(1). node in the case of making location error as small as possible.
International Journal of Distributed Sensor Networks 11

D
P2

A P1
dA1
A
dB1 P1
O
B
dC1

C C
B

Figure 13: Unknown nodes of P1 and P2 for locating.

Figure 12: A set of RSSI vector is uniquely correspondent to the

coordinates of node.
There is a unique solution for the coordinate of P1 by above
formula (43) and formula (44).
There is a known assumption that
3.2.1. The Determination of the Location Unit
𝑑𝐴1 = 𝑑𝐴2 ,
Theorem 4. The midpoint of the each edge of the triangle is
taken; the new three midpoint points and three top points of 𝑑𝐵1 = 𝑑𝐵2 , (39)
the original triangle are regarded as reference sample points;
there is uniqueness of the RSSI vector which is formed by the 𝑑𝐶1 = 𝑑𝐶2 .
RSSI value of sample points which are received from all anchor
nodes. So, there is the same coordinates of P1 and P2; clearly, there is
a contradiction relationship with known assumptions, so the
Proof (by reduction). As shown in Figure 12, it is assumed that
original theorem is established. That has just finished.
there are different coordinates of two different nodes P1 and
P2, but they have the same RSSI vector with same dimension
and value (the numbers of anchor nodes are 3, the dimension According to Theorem 4, there is uniqueness for every
of vector is 3), 𝑃1(𝑥, 𝑦), the distance from P1 to anchor nodes vector of reference sample points in subsequent localization
A, B, and 𝐶 are 𝑑𝐴1 , 𝑑𝐵1 , and 𝑑𝐶1 ; similarly, in terms of P2, algorithm.
distances are 𝑑𝐴2 , 𝑑𝐵2 , and 𝑑𝐶2 .
3.2.2. The Main Idea of the Localization Algorithm. As shown
∵ There is the same RSSI vector with same dimension in Figure 13, P1 and P2 are indicated unknown nodes, A, B,
and valuefor P1 and P2 C, and 𝐷 are indicated reference anchor nodes which are
the closest to P1 and P2, and the quadrilateral indicated the
⇒ By Shadowing model, there is a same distance location unit, which is surrounded by A, B, C, and D. AC
vector with same dimension and value for P1 and P2. and BD indicated two diagonals of the quadrilateral ABCD
and 𝑂 indicated diagonals intersection (coordinates can be
⇒There is the same distance from P1 or P2 to any obtained); it is regarded as the new reference sample point.
same anchor nodes. By the following relation of area constraint, the location
of unknown node can be determined roughly as
∴ There is the following equation:
𝑆𝐴𝐵𝐶𝐷 = 𝑆Δ𝐴𝐵𝑃 + 𝑆Δ𝐵𝐶𝑃 + 𝑆Δ𝐶𝐷𝑃 + 𝑆Δ𝐷𝐴𝑃 . (40)
2 2
√(𝑥 − 𝑥𝐴 ) + (𝑦 − 𝑦𝐴 ) = 𝑑𝐴1 ,
⇒Point 𝑃 is inside of quadrilateral of ABCD
√(𝑥 − 𝑥𝐵 )2 + (𝑦 − 𝑦𝐵 )2 = 𝑑𝐵1 ,
𝑆Δ𝐷𝐴𝑂 = 𝑆Δ𝐷𝑃𝐴 + 𝑆Δ𝐴𝑃𝑂 + 𝑆Δ𝑂𝑃𝐷. (41)
√(𝑥 − 𝑥𝐶)2 + (𝑦 − 𝑦𝐶)2 = 𝑑𝐶1 ,
(38) ⇒Point 𝑃 is inside of Δ𝐷𝐴𝑂.
−1 By Algorithm 1 of two location system, the location of
𝑥 2 (𝑥𝐴 − 𝑥𝐶) 2 (𝑦𝐴 − 𝑦𝐶)
( )=( ) unknown node can be determined.
𝑦 2 (𝑥𝐵 − 𝑥𝐶) 2 (𝑦𝐵 − 𝑦𝐶)

𝑥2 − 𝑥𝐶2 + 𝑦𝐴2 − 𝑦𝐶2 + 𝑑𝐶2 − 𝑑𝐴

2
3.2.3. The Location Mechanism of the Unknown Node Is inside
× ( 𝐴2 2 2 2 2 2 ).
𝑥𝐵 − 𝑥𝐶 + 𝑦𝐵 − 𝑦𝐶 + 𝑑𝐶 − 𝑑𝐵 of the Location Unit. As shown in Figure 14, if the unknown
12 International Journal of Distributed Sensor Networks

The input: These reference anchor nodes 𝐴, 𝐵, 𝐶, 𝐷, which are the closest to 𝑃𝑖 , they are made
of a set 𝐶𝑖 : 𝐶𝑖 = {𝐶𝑖𝐴 , 𝐶𝑖𝐵 , 𝐶𝑖𝐶, 𝐶𝑖𝐷 }.
There are number Un unknown nodes are distributed in the region of locating.
The output: The unknown node 𝑃 is inside of the location unit or external.
Step 1. FOR 𝑖 = 1: Un
Step 2. IF 𝑆𝐴𝐵𝐶𝐷 = 𝑆Δ𝐴𝐵𝑃 + 𝑆Δ𝐵𝐶𝑃 + 𝑆Δ𝐶𝐷𝑃 + 𝑆Δ𝐷𝐴𝑃
Step 3. RUN InternA
//The algorithm of location mechanism that 𝑃 is inside of graphics.
Step 4. ELSE
Step 5. RUN ExternA
//The algorithm of location mechanism that 𝑃 is outside of graphics.
Step 6. ENDIF
Step 7. ENDFOR

Algorithm 1: Decide node in internal or external.

D P D
H
N
E L O2
J O2 A
P
M
II A P1
A G
P2
O
F
O

C
B
C
B
Figure 15: The unknown node is outside of the graph.
Figure 14: The division of the location unit when the unknown node
is inside of graphics and the determination of sample points.

node is inside of ABCD, some operations will be taken to as shown in Figure 15, the coordinates of the unknown node
reduce the location unit as follows. is determined by determining two triangles of copoint.
Main operations are as follows.
Step 1. Point 𝑃 is judged whether in Δ𝐴𝐵𝑂, Δ𝐵𝐶𝑂, Δ𝐶𝐷𝑂,
and Δ𝐷𝐴𝑂 or not. Step 1. Points 𝐷 and 𝐴 are found, which are first and second
of the RSSI vector of the unknown node 𝑃; they are made of
Step 2. If point 𝑃 is inside of any one triangle, three new Δ𝑃𝐴𝐷.
reference sample points are obtained by taking the middle
point of each edge of triangle. Step 2. Points 𝐷 and 𝐵 are found, which are first and second
of the RSSI vector of the unknown node 𝑃; they are made of
Step 3. The similar degree for the RSSI vector of the unknown Δ𝑃𝐵𝐷.
node 𝑃 are compared with the top points of the original
triangle and the RSSI vector of three new reference sample Step 3. Since the RSSI can be measured, the distance between
points (there are six RSSI vectors); the most similar reference points can be obtained by the signal attenuation model.
sample points are found; namely, three reference sample Since three lengths of sides are known, the triangle area can
points which are closest to 𝑃 are 𝐸, 𝐺, 𝐷. be obtained. Since coordinates of two tops are known, the
By analogy, above steps are repeated, reference sample height from point 𝑃 to its edge can be obtained by the area
points are obtained by looking for the midpoint of the formulation 𝑆 = (1/2)𝑎 × ℎ.
location triangle constantly, which are the closest to the The coordinate of known node is obtained by the follow-
unknown node P, the microtriangle region which included ing equations:
the unknown node is narrowed, and Figure 14 has shown that
midpoint 𝑃 is locked in the region of Δ𝐿𝑁𝑀. 𝐿 = 𝑑𝐴𝐷 + 𝑑𝐷𝑃 + 𝑑𝑃𝐴 ,

3.2.4. The Location Mechanism of Unknown Node Is outside of 𝐿 𝐿 𝐿 𝐿

the Location Unit. If the unknown node is outside of the unit, 𝑆Δ𝑃𝐴𝐷 = √ ∗ ( − 𝑑𝐴𝐷) ∗ ( − 𝑑𝐷𝑃 ) ∗ ( − 𝑑𝑃𝐴 ),
2 2 2 2
International Journal of Distributed Sensor Networks 13

1 P1󳰀
𝑆Δ𝑃𝐴𝐷 = 𝑑𝐴𝐷 ∗ 𝑑𝑃𝑂1 , P2 󳰀
2 D

𝐴∗𝑥+𝐵∗𝑦+𝐶
𝑑𝑃𝑂1 = . O1
√𝐴2 + 𝐵2 O2
A
(42)
A
P1
In this formula, 𝐿 indicated the circumference of the triangle, P2
𝑆 indicated the area of the triangle, 𝑑 indicated length of sides O
of the triangle, 𝐴 ∗ 𝑥 + 𝐵 ∗ 𝑦 + 𝐶 = 0 indicated the linear
equation which is included the line 𝐴𝐷, and two groups of
coordinates of unknown nodes can be obtained by (47). C
By the same token, RSSI values of points 𝐷 and 𝐵 are B
chosen, which are first and third in RSSI vector. Δ𝑃𝐵𝐷 is Figure 16: The complementary localization algorithm used the
made of points 𝑃, 𝐷, and 𝐵; two groups of coordinates generalized inverse.
of unknown nodes can be obtained; they are 𝑃(𝑋, 𝑌) and
𝑃1(𝑋2, 𝑌2). The public top point 𝑃 of two triangles can
be obtained by synthesizing two groups of solutions; the ̃ = 𝐺𝑏 indicated
𝐺 indicated a matrix, 𝐺 = 𝐴−𝑖 , 𝑋
coordinate of the unknown node is 𝑃(𝑋, 𝑌).
the least squares solution of incompatible equations, and
Although this method is feasible, the rate of measurement
‖𝐴𝐺𝑏 − 𝑏‖2 is minimum, so ‖𝐴(𝐺𝑏 + (𝐼 − 𝐺𝐴)𝑍) − 𝑏‖2 =
error of RSSI on all directions may not be consistent in the
‖𝐴𝐺𝑏 − 𝑏‖2
process of distance measuring; it will be leaded to distance ̃ = 𝐴− 𝑏 and 𝑋 ̃ = (𝐺𝑏 + (𝐼 − 𝐺𝐴)𝑍) both are
changes of the corresponding direction showing different Namely, 𝑋 𝑖
conditions; at this time, the above method cannot get two least squares solutions of incompatible linear equations.
public point of two triangles; namely, there is not public The coordinate of the unknown node is calculated accord-
solution for simultaneous equations. ing to anchor nodes of A, B, and D, and there are equations
as follows:
󵄨󵄨 󵄨
3.2.5. The Method of Generalized Inverse. As shown in 󵄨󵄨𝑘𝐴𝐷 ∗ 𝑥 − 𝑦󵄨󵄨󵄨 + 𝑏1 − ℎ𝐴𝐷⊥ ∗ √𝑘𝐴𝐷
2 + 1 = 0,
Figure 16, there are random measurement errors of RSSI; the
calculation result is not coincidence. Furthermore, there is 󵄨󵄨 󵄨
not public solution for simultaneous equations. To this end, a 󵄨󵄨𝑘𝐵𝐷 ∗ 𝑥 − 𝑦󵄨󵄨󵄨 + 𝑏2 − ℎ𝐵𝐷⊥ ∗ √𝑘𝐵𝐷
2 + 1 = 0,

complementary localization algorithm is designed when the

2 (𝑥𝐴 − 𝑥𝐷) 𝑥 + 2 (𝑦𝐴 − 𝑦𝐷) 𝑦
node is outside of the location unit by using the generalized (44)
inverse [36]. = 2
𝑑𝑃𝐴 − 2
𝑑𝑃𝐷 + 2
𝑥𝑃𝐷 − 2
𝑥𝑃𝐴 + 2
𝑦𝑃𝐷 − 2
𝑦𝑃𝐴 ,
𝐴 ∈ 𝐶𝑚×𝑛 and 𝑏 ∈ 𝐶𝑚 are both known; the vector 𝑋 ∈ 𝐶𝑚
is unknown, which is included in the linear equations 𝐴𝑋 = 2 (𝑥𝐵 − 𝑥𝐷) 𝑥 + 2 (𝑦𝐵 − 𝑦𝐷) 𝑦
𝑏.
2 2 2 2 2 2
If there is a solution of equations, the equation is = 𝑑𝑃𝐵 − 𝑑𝑃𝐷 + 𝑥𝑃𝐷 − 𝑥𝑃𝐵 + 𝑦𝑃𝐷 − 𝑦𝑃𝐵 .
regarded as the compatibility equations; on the contrary, it
is the incompatible equation. In terms of the incompatible Above equation set is turned into a form such as the linear
equations, there is no general solution. So in this section, equations 𝐴𝑋 = 𝑏; the coefficient matrix 𝐴 and vector 𝑏 are
the optimal and approximate solution of incompatible and as follows:
insoluble equations is obtained by using the generalized
inverse 𝑘𝐴𝐷 −1
[ 𝑘 −1 ]
𝐴=[ 𝐵𝐷 ]
[2 (𝑥𝐴 − 𝑥𝐷) 2 (𝑦𝐴 − 𝑦𝐷)] ,
𝑋 = min𝑚 ‖𝐴𝑋 − 𝑏‖ . (43)
𝑋∈𝐶
[ 2 (𝑥𝐵 − 𝑥𝐷) 2 (𝑦𝐵 − 𝑦𝐷) ]
2 +1−𝑏
(45)
The solution of the above formula is regarded as the least ℎ𝐴𝐷⊥ ∗ √𝑘𝐴𝐷 1
squares solution of incompatible equations; the symbol ‖ ‖ is [ ]
[ 2 ]
the European norm. 𝑏=[ ℎ𝐵𝐷⊥ ∗ √𝑘𝐵𝐷 + 1 − 𝑏2 ].
[ 2 ]
[𝑑 − 𝑑2 + 𝑥2 − 𝑥2 + 𝑦2 − 𝑦2 ]
𝑃𝐴 𝑃𝐷 𝑃𝐷 𝑃𝐴 𝑃𝐷 𝑃𝐴
Definition 5. If the incompatible linear equation 𝐴𝑋 = 𝑏 2 2 2 2 2 2
[ 𝑑𝑃𝐵 − 𝑑𝑃𝐷 + 𝑥𝑃𝐷 − 𝑥𝑃𝐵 + 𝑦𝑃𝐷 − 𝑦𝑃𝐵 ]
is satisfied with 𝑋 = 𝐴−𝑖 𝑏, 𝐴−𝑖 indicated the least square
generalized inverse of 𝐴; there will be a relation that 𝑋 ̃ Because there is no common solution of equations, it is
is satisfied with ‖𝐴𝑋 ̃ − 𝑏‖2 < ‖𝐴𝑋 − 𝑏‖2 , so 𝑋 ̃ = 𝐴 𝑏 − the incompatible linear equations. Then, the least square
𝑖
indicated the least square solution of equations, and there is generalized inverse of equations is as follows:
the minimum sum of squares of error by comparing to other −1
solutions. 𝐴−𝑖 = (𝐴𝑇 𝐴) 𝐴𝑇 . (46)
14 International Journal of Distributed Sensor Networks

10 0.7
9
0.6
8
0.5

Localization error (m)

7
6
0.4
5
4 0.3

3 0.2
2
0.1
1
0 0
0 2 4 6 8 10
8 16 24 32
Anchor number
Unknown node
Anchor node Figure 18: The relation between the number of anchor nodes and
location error.
Figure 17: The layout of anchor nodes and unknown nodes.

0.7
So, the least squares solution of incompatible linear equations
is as follows: 0.6

̃ = 𝐴− 𝑏
𝑋 𝑖
Localization error (m)

0.5

2 +1−𝑏
ℎ𝐴𝐷⊥ ∗ √𝑘𝐴𝐷 1 0.4
[ ]
[ 2 ] (47)
−[ ℎ ∗ √𝑘𝐵𝐷 + 1 − 𝑏2 ]
= 𝐴𝑖 [ 𝐵𝐷⊥ ]. 0.3
[ 2 2 2 2 2 2 ]
[𝑑𝑃𝐴 − 𝑑𝑃𝐷 + 𝑥𝑃𝐷 − 𝑥𝑃𝐴 + 𝑦𝑃𝐷 − 𝑦𝑃𝐴 ]
2 2 2 2 2 2 0.2
[ 𝑑𝑃𝐵 − 𝑑𝑃𝐷 + 𝑥𝑃𝐷 − 𝑥𝑃𝐵 + 𝑦𝑃𝐷 − 𝑦𝑃𝐵 ]
̃ = [ 𝑦𝑥̃̃ ]; the estimation coordinate of the
Consider 𝑋 0.1
̃ 𝑦).
unknown node is 𝑃(𝑥, ̃
0
RSSI mea error = 0.2 RSSI mea error = 0.1
4. Results of Simulation Experiment and RSSI measurement error
Real Experiment N=8 N = 24
N = 16 N = 32
4.1. The Result of Simulation Experiment. In order to simulate
the real environment, the Shadowing model is used to convert Figure 19: The relation between RSSI ranging error and location
RSSI value into its corresponding distance; ranging errors error.
are setup to simulate the range environment. In this section,
localization algorithm is simulated by the tool of MATLAB
7.0, 8 anchor nodes and 160 unknown nodes are taken in this In order to describe the effection of ranging error and
experiment, all nodes are isomorphic, unknown nodes are the trend of ranging error, we have drawn a picture of the
random distributed in the rectangle area 10 × 10 m2 in the relationship between location error and ranging error. In
experiment environment, and anchor nodes are in the scope Figure 19, the increasing trends of anchor nodes are increased
of the region that unknown nodes can be communicated from 8 to 32, and the changing of the location error is
(Figure 17). changed.
Figure 18 has shown that location error is obtained by At the same time, under the condition that the total
simulation when anchor nodes numbers are 8, 16, 24, and number of unknown nodes is invariable at UN = 160,
32. Along with the increasing numbers of anchor nodes, Figure 19 has shown the simulation result when RSSI ranging
location errors are decreased gradually. Because the more the errors are 0.1 and 0.2, which are obtained by the method of
numbers of anchor nodes are in the locating area, the more average calculation, and it also means that the percentages of
the numbers of anchor nodes are nearest to the unknown RSSI ranging error are 10% and 20%. According to Figure 19,
node, the smaller quadrilateral area is formed by anchor when the number of anchor nodes is unchanged, the bigger
nodes; namely, sample reference points are more closer to the the RSSI ranging error, the larger the location error; at the
unknown node. same time, when the RSSI ranging error is constant, the more
International Journal of Distributed Sensor Networks 15

Table 3: The relationship of the number of reference sample Localization error and average error: N = 16
4
points between localization algorithm in this paper and localization
algorithm of LTFM model. 3.5
Algorithm S/64 S/256 S/512
3
Our algorithm 13 16 19

Location error (m)

FTLM model 13 25 37 2.5

2
Localization error and average error: N = 8
4
1.5
3.5
1
3
0.5
Location error (m)

2.5
0
0 20 40 60 80 100 120 140 160
2 Unknown node number

1.5 L error Dis mea error = 0.1

Dis mea error = 0.05 Dis mea error = 0.2
1
Figure 21: The number of anchor nodes is 16.
0.5

0 Localization error and average error: N = 24

0 20 40 60 80 100 120 140 160 4
Unknown node number
3.5
L error mean L error = 0.1
mean L error = 0.05 mean L error = 0.2 3
Location error (m)

Figure 20: The number of anchor nodes is 8. 2.5

the number of anchor nodes, the smaller the location error. It 1.5
is also suggested that location error based on RSSI is not only
dependent on the localization algorithm, but also dependent 1
on the accuracy of ranging analysis of RSSI [10].
0.5
There is a relation analysis between the number of anchor
nodes and location error and RSSI range error in Figure 19, in 0
the real measurement environment. The measurement errors 0 20 40 60 80 100 120 140 160
of RSSI at all directions are not consistent; the generalized Unknown node number
inverse is used to solve the problem that there is not public L error Dis mea error = 0.1
solution of equations which is formed by the external location Dis mea error = 0.05 Dis mea error = 0.2
unit. The picture of location error is shown from Figures
20, 21, 22, and 23; the colorful lines indicated the mean Figure 22: The number of anchor nodes is 24.
location error when RSSI ranging errors are 0.05, 0.1, and
0.2, according to these figures, the mean location error =
0.2 is regarded as the better mean value, along with the When the location area is 10 × 10 m2 , the comparison of
increasing number of anchor nodes; location error has shown location error between the new algorithm and the sequence
a decreasing trend on the whole; in addition, the final location three-orthocenter method and the sequence location method
error of the random ranging error is almost constant with the and three other center methods is shown, respectively, in
final location error of the constant ranging error before, but Figure 24.
it is practical in application. Visibly, along with the decreasing number of anchor
In Table 3, there is a comparative relationship of the num- nodes, location error of these methods will be increased,
ber of reference sample points required when each unknown because the more the number of anchor nodes, the more
node is required to locate between localization algorithm and the times that original location area is divided into different
localization algorithm of FTLM model. According to Table 3, formulas of smaller area, unknown node is more closer to
it can be seen that this algorithm needed less reference sample anchor nodes, the more accurate location; Figure 24 has
points when the microarea of positing is the same one; shown that location error of three-orthocenter method is
moreover, less amount of calculation is needed. The iteration more than five times than new location algorithm when the
location by reducing the area is convergent. anchor node number is 8, because the new location algorithm
16 International Journal of Distributed Sensor Networks

Table 4: Location error of unknown nodes.

Unknown node
Number 1 Number 2 Number 3 Number 4
Location error (m) 0.24 0.36 0.37 0.10

Localization error and average error: N = 32 3

4

3.5 2.5

3
2
Location error (m)

2.5
1.5
2

1.5 1

1
0.5
0.5
0
0 0 0.5 1 1.5 2 2.5 3
0 20 40 60 80 100 120 140 160
Unknown node number Anchor node
Real coordinate
L error Dis mea error = 0.1 Estimated coordinate
Dis mea error = 0.05 Dis mea error = 0.2
Figure 23: The number of anchor nodes is 32. Figure 25: The result of location.

3 sequence three-orthocenter method and sequence location

method are already increased to 462273 when the anchor
2.5 node number is increased to 32; the area is quite small,
which included the unknown node, but, obviously, lots of
Location error (m)

2 measurements should be taken; a great amount of calculation

and sequence alignment should be made; at this point, the
16 reference sample points are needed to compare with each
1.5
unknown node in the new location algorithm; the location
error is lower than three-orthocenter method; the superiority
1 is shown.
In this simulation experiment, four anchor nodes and
0.5 four unknown nodes are set up in the region 3 × 3 m2 .
The mean absolute ranging error of nodes is obtained as
0 0.27 m; the percentage of location error is about 17.6% by the
10 15 20 25 30 calculation formula:
Anchor node number
󵄨󵄨 󵄨
󵄨𝜎 󵄨󵄨
Our algorithm 𝑃𝑖 = 󵄨󵄨 󵄨 𝑖−𝑥 󵄨 󵄨󵄨 × 100%. (48)
Three centroid in triangle with sequence algorithm 󵄨󵄨𝑑𝑖 − 𝜎𝑖−𝑥 󵄨󵄨
SBL
Three centroid in triangle In this formula, 𝑃𝑖 indicated the percent of location error of
the 𝑖th unknown node, 𝜎𝑖−𝑥 indicated the absolute ranging
Figure 24: The comparison between algorithms.
error between real coordinate of the 𝑖th unknown node and
estimated coordinate, 𝑑𝑖 indicated the distance between the
𝑖th unknown node and original point.
continuously narrowed the area which included the unknown In Figure 25, black points indicated anchor nodes which
node by the scale of 1/4 in the process of location; thus the are settled before, red points indicated the real coordinates of
actual location of the original node can be better restored; in unknown nodes (they are ranged by tools), and blue triangle
this paper, the location error of the new location algorithm is indicated estimated coordinates of unknown nodes by the
almost constant with the sequence three-orthocenter method new location algorithm.
and the sequence location method when the anchor node The location error is not only related to the location
number is 32, because the numbers of location sequence of algorithm but also related to RSSI range and the number of
International Journal of Distributed Sensor Networks 17

Table 5: 20 groups of value of signal strength. Table 6: The measured data of number 1 node.

Node 1 Node 2 Node 3 Node 4 Number Distance Value Count Percent

1m 4m 1m 4m 1m 4m 1m 4m 67 4 1.00%
98 33 106 50 106 70 123 70 72 80 20.00%
98 33 106 50 103 70 123 70 1m 75 8 2.00%
100 25 106 50 106 70 123 70 78 10 2.50%
100 25 106 47 106 70 123 70 81 20 5.00%
98 25 106 50 103 56 123 61 1 50 4 1.00%
98 25 106 53 103 56 123 61 58 21 5.25%
95 22 106 44 103 64 123 61 64 37 9.25%
4m
95 25 106 44 103 64 123 61 67 70 17.50%
98 19 106 44 103 64 123 61 70 121 30.25%
98 19 106 44 103 64 123 61 72 147 36.75%
98 19 106 44 103 64 123 67
98 19 106 44 103 64 123 67 Table 7: The measured data of number 2 node.
98 19 106 44 103 61 123 67
Number Distance Value Count Percent
98 19 106 44 103 61 123 64
100 15 3.75%
98 19 106 44 106 61 123 64
103 81 20.25%
98 19 106 47 106 61 123 64
1m 106 300 75.00%
98 19 106 47 106 58 123 64
109 4 1.00%
98 19 106 47 106 58 123 64
36 9 2.25%
95 19 106 47 106 58 123 64
2 39 48 12.00%
95 19 106 50 106 58 123 64
42 20 5.00%
44 52 13.00%
4m
47 25 6.25%
anchors nodes. There is an emulational result of location error 50 129 32.25%
of these unknown nodes shown as in Table 4. 53 82 20.50%

4.2. The Result of Real Experiment. In this experiment, the Table 8: The measured data of number 3 node.
latticed experiment environment is designed; there are some
nodes in it. Because the place of this experiment is in Number Distance Value Count Percent
laboratory, the location grid 20 × 20 m2 is set; 4 × 4 m2 is just a 14 16 4%
part of the location grid. In order to choose anchor nodes with 70 36 9.00%
better performance, the amount of calculation is needed to 1m 72 21 5.25%
decrease; four anchor nodes and one known node are placed 75 15 3.75%
in the location grid 4 × 4 m2 . 78 81 20.25%
In addition, the unknown node is moved at different 3 81 67 16.75%
places. First of all, RSSI values of four nodes in multiple 86 91 22.75%
directions are measured, respectively; there is a result that 89 32 8.00%
RSSI values of nodes in each direction are basic stability when 4m
92 19 4.75%
they are individual, so nodes are required to be immobilized.
95 22 5.50%
Secondly, the range models of nodes in different directions
are determined, we found that location errors are unreal by 81 86 21.50%
calculating, so we measured the RSSI value of four nodes in
the range of 1 m and 4 m, respectively, and the total RSSI value
of are more than 2000 groups. Thirdly, the statistic of the is needed to build. There are 20 groups of RSSI value of four
probability of each RSSI value is obtained by using the tool of nodes in the range of 1 m and 4 m, respectively, in Table 5.
Matlab; we selected the optimal value by using the Gaussian The number of nodes and distance of nodes and the prob-
fitting. Finally we built the ranging model by introducing ability of each RSSI value are measured in the experiment. For
𝑑 = RSSI𝑖 ∗ 𝑘 + 𝑏. example, there are parts of information from Tables 6, 7, 8,
In terms of the station that the increasing anchor nodes and 9.
in the experiment, the range model of this node should be The optimal value is selected by using the Gaussian fitting.
created. Firstly, because interference factors are inevitable Figures 26, 27, 28, 29, 30, 31, 32, and 33 have shown curves of
indoor, there are different RSSI value of anchor nodes in a Gaussian fitting of four nodes when their distances are 1 m
different direction. Secondly, the new ranging model of node and 4 m.
18 International Journal of Distributed Sensor Networks

Table 9: The measured data of number 4 node. Table 12: Coordinates of four nodes.
Number Distance Value Count Percent Number
100 4 1.00% NO. 1 NO. 2
103 182 40.50% Real coordinate (0.00, 0.00) (1.00, 0.50)
1m 106 50 12.50% Mean coordinate (0.07, −0.50) (0.62, −1.00)
123 164 41.00% Maximum (m) 0.72 1.87
61 75 18.75% Minimum (m) 0.40 1.31
4 64 79 19.75% Mean (m) 0.52 1.53
67 127 31.75% Number
70 59 14.75% NO. 3 NO. 4
4m
72 60 15.00% Real coordinate (−1.00, 0.00) (0.05, −1.00)
100 4 1.00% Mean coordinate (−0.80, 0.99) (−0.10, −1.70)
103 182 40.50% Maximum (m) 1.61 1.84
Minimum (m) 0.59 0.48
Table 10: Range models of nodes. Mean (m) 1.04 1.02

Number Range models

1 𝑑 = RSSI𝑖 ∗ (−0.0421) + 4.9995
2 𝑑 = RSSI𝑖 ∗ (−0.0552) + 6.8070
0.6
3 𝑑 = RSSI𝑖 ∗ (−0.0764) + 8.9277
4 𝑑 = RSSI𝑖 ∗ (−0.0541) + 7.5677 0.5

Table 11: Coordinates of unknown nodes. 0.4

Real coordinate (0.00, 0.00) (1.00, 0.50) 0.3

(0.02, −0.40) (0.48, −0.70)
(0.00, −0.60) (0.38, −0.80) 0.2
(0.00, −0.40) (0.34, −0.80)
(0.00, −0.60) (0.48, −0.70) 0.1

(−0.10, −0.60) (0.62, −1.00)

Measured coordinate 0
(0.00, 0.72) (0.48, −1.30) 70 75 80 85 90 95
(0.00, −0.40) (0.62, −1.00)
(0.02, −0.40) (0.48, −1.00) y vs. x
Fit 1
(0.07, −0.50) (0.62, −1.00)
(0.02, −0.60) (0.62, −1.00) Figure 26: The fitting curve of number 1 node when distance is 1 m.
Real coordinate (−1.00, 0.00) (0.50, −1.00)
(−0.80, 0.99) (0.03, −0.70)
(−0.80, 0.55) (−0.50, −1.60)
(−0.80, 0.99) (−0.80, −2.20) is obtained from the noninterfering environment. For the
(−0.80, −1.60) (−0.80, −1.80) first time, the smallest location error is 0.4 m in the actual
(−0.80, 1.51) (−0.30, −1.10)
coordinates of node; for the second time, the smallest location
Measurement coordinate error is 1.31 m; for the third time, the smallest location error
(−0.80, −1.20) (−0.30, −1.40)
is 0.59 m; for the fourth time, the smallest location error is
(−0.70, −1.30) (−0.20, −1.20)
0.48 m (see Table 12).
(−0.70, 0.90) (−0.20, −0.90) In Table 9, since interference factors are inevitable indoor,
(−0.70, 1.13) (0.00, −1.60) there are different RSSI values of anchor nodes in different
(−0.70, 0.90) (−0.10, −1.70) direction. Their ranging models are based on different nodes.
There are different results of mean location error.
In Figure 34, black squares indicated fixed anchor nodes
The piecewise linear interpolation for RSSI-d curve is in the experiments, yellow dots indicated actual coordinates
used to ensure the ranging precision. There are four range of unknown nodes, and blue triangle points indicated esti-
models of four nodes as in Table 10. mation coordinates of unknown nodes which are obtained
There are real coordinates of nodes. There is part of by the new location algorithm. In each experiment, the
measurement data in Table 11. mean location error is about 1 m. Four yellow dots indicated
Through times of measurement of the unknown node, the movement results; there are four times of movement for the
smallest location error is regarded as the ideal date which unknown node; the experiment is regarded as a dynamic test.
International Journal of Distributed Sensor Networks 19

0.3
0.3

0.25
0.25

0.2
0.2

0.15 0.15

0.1 0.1

0.05 0.05

0 0
20 22 24 26 28 30 32 34 36 40 45 50 55

y vs. x y vs. x
Fit 1 Fit 1

Figure 27: The fitting curve of number 1 node when distance is 4 m. Figure 29: The fitting curve of number 2 node when distance is 4 m.

0.9
0.55
0.8
0.5
0.7
0.45
0.6
0.4
0.5
0.35
0.4
0.3
0.3
0.25
0.2
0.2
0.1
0.15
0
100 101 102 103 104 105 106 107 108 109 100 101 102 103 104 105 106

y vs. x y vs. x
Fit 1 Fit 1

Figure 28: The fitting curve of number 2 node when distance is 1 m. Figure 30: The fitting curve of number 3 node when distance is 1 m.

4.3. The Simulated Application of the Location Algorithm 4.3.2. The Design of Mobile Terminal. This system can be
applied to prevent children lost indoor; accounting that the
4.3.1. The Design of Location System. The location system children’s speed of movement is relatively much slower than
is applied at home and location area is set as 4 × 4 m2 . parents, the process of implement is regarded as a lower level
Four anchor nodes are It is selected in the location region. dynamic environment. The system is able to work well when
A RFID tag is carried with robot which is regarded as the
child moves into room.
unknown node. In addition, the location system is set as a
server and a transmission PC and a router (see Figure 35). There are some pictures of experiment displayed in the
Firstly, anchor nodes send strength signal to the robot which interface of mobile terminal. In Figure 36, the application is
can be received and sent signal. Secondly, the server received set at home, so the background of the interface is the picture
signal information from the robot and processed these data. of home, and the black point indicated the mobile robot.
Thirdly, mobile terminal gets processed data from server by In Figure 37, it is shown that there is an alarm’s tooltip
router. Finally, information of mobile robot is displayed in the in the picture; it indicated that the RFID tag is read by
interface of mobile terminal. Particularly, the card reader is card reader; namely, the mobile robot is outside of home.
set at the door; it is a caution of the mobile robot when the Particularly, this system designed the alarm’s music with the
robot is outside of the location region. alarm’s function.
20 International Journal of Distributed Sensor Networks

0.3
0.2
0.28

0.26
0.15
0.24

0.22
0.1
0.2

0.18
0.05
0.16

0.14
0
20 25 30 35 40 45 50 55 60 65 70 62 64 66 68 70 72

y vs. x y vs. x
Fit 1 Fit 1

Figure 31: The fitting curve of number 3 node when distance is 4 m. Figure 33: The fitting curve of number 4 node when distance is 4 m.

0.45
2.1
0.4
1.7
0.35

0.3
1.3
0.25
0.9
0.2

0.15 (−0.8, 0.55)

0.5 (1, 0.5)
0.1

0.05 0.1
(−1, 0) (0, 0)

100 105 110 115 120 −0.3

(0, −0.4)
y vs. x (0.04, −0.7)
.
−0.7 (0.48, −0.7)
Fit 1
−0.9
Y (m) −1.5 −1.1 −0.7 −0.3 0.1 0.5 0.9 1.3 1.5
Figure 32: The fitting curve of number 4 node when distance is 1 m.
x (m)

Figure 34: The effect of location.

5. Conclusion
In this paper, we studied on the location technology of RSSI-
based in WSN. In order to get accurate RSSI value and
simulate the real experimental environment, there is a large Particularly, the generalized inverse is solved to estimate
amount of experimental data. In the process of location, the coordinate of unknown node by equations. Simulation
we determined the related parameters of the Shadowing and experiment results show that our approach outperforms
model and analyzed the RSSI value at the specific distance existing approaches in terms of location accuracy, and our
by the Gaussian fitting; in addition, not only we built location system is applied to locate the robot quite well.
the RSSI-based interpolation model of node, but also we
analyzed the influence of the number of nodes. In terms
of the new location algorithm, we proposed the location Conflict of Interests
mechanism; it is used to estimate that the unknown node is
internal or external. Moreover, we proposed the concept of The authors declare that there is no conflict of interests
vector similar degree; it is helpful to choose anchor nodes. regarding the publication of this paper.
International Journal of Distributed Sensor Networks 21

Acknowledgments
Node Door
The authors would like to thank the Chongqing Natural
Reader Science Foundation under Grant no. cstc2012jjA40038 and
Tag
the Science and Technology Research Project of Chongqing
Mobile Municipal Education Commission of China. The work pre-
terminal
Robot sented in this paper was supported in part by the Ministry of
Industry and Information Technology of China for the special
Host
funds of Development of the Internet of Things (2012-583).
Router

References
[1] R. Lin, Z. Wang, R. Sun, and L. Sun, “Vision-based mobile robot
localization and mapping using the PLOT features,” in Proceed-
ings of the IEEE International Conference on Mechatronics and
Automation (ICMA ’12), pp. 1921–1927, Chengdu, China, August
Server 2012.
[2] S.-B. Han, J.-H. Kim, and H. Myung, “Landmark-based particle
localization algorithm for mobile robots with a fish-eye vision
Figure 35: The schematic diagram of the location system. system,” IEEE/ASME Transactions on Mechatronics, vol. 18, no.
6, pp. 1745–1756, 2013.
[3] H. Zhang, J. Chen, and K. Zhang, Reliable and Efficient RFID-
Based Localization for Mobile Robot, IEEE, Washington, DC,
Living Kitchen USA, 2013.
room Balcony [4] F. J. Shang, Communication Protocol in Wireless Sensor Net-
Dining
room works, Electronic Industry Press, Beijing, China, 2011.
[5] J. Z. Li and B. J. Li, “The concept and problem and progress of
Door
data management in Sensor Networks,” Journal of Software, vol.
Study Lounge 14, no. 10, pp. 1717–1727, 2003.
Cloakroom
[6] L. M. Sun, Wireless Sensor Network, Tsinghua University Press,
Bedroom
Bathroom Beijing, China, 2005.
Tearoom [7] J.-Y. Huang and C. H. Tsai, “Improve GPS positioning accuracy
with context awareness,” in Proceedings of the 1st IEEE Interna-
Bathroom tional Conference on Ubi-Media Computing and Workshops, pp.
Living Bedroom 94–99, August 2008.
Gymnasium room
[8] B. Peng, “Research on compensatory location algorithm based
on RSSI ranging error in Wireless Sensor Network,” Chinese
Journal of Dalian Technology University. In Press.
Figure 36: The normal effect.
[9] Y. Guan Wang and Z. Bian Wang, Wireless Sensor Networks,
Electronic Industry Press, Beijing, China, 2012.
[10] Y. D. Dong, G. D. Yuan, and F. X. Jie, “An indoor location
Kitchen algorithm base on RSSI-Similarity degree,” Chinese Journal of
Living
room Balcony Sensors and Actuators, vol. 22, no. 2, pp. 264–268, 2009.
Dining [11] H. Woo, C. Lee, and S. Oh, “Reliable anchor node based
room
range-free localization algorithm in anisotropic wireless sensor
networks,” in Proceedings of the 27th International Conference on
Door
Information Networking (ICOIN ’13), pp. 618–622, January 2013.
Study [12] M. Di, E. M. Joo, W. Bang, and L. H. Beng, “Range-free
Lounge Cloakroom localization based on hop-count quantization in wireless sensor
Bedroom
networks,” in Proceedings of the IEEE Region Conference, pp. 1–6,
Bathroom November 2009.
Tearoom [13] T. Borangiu and I. Stanculeanu, “Enhanced RSSI localization
system for asset tracking services using Non expensive IMU,”
in Proceedings of the 14th IFAC Symposium on Information
Bathroom Control Problems in Manufacturing (INCOM '12), pp. 1838–1843,
Warning! Bedroom Bucharest, Romania, May 2012.
Gymnasium [14] R. M. Rahman, K. Barker, and R. Alhajj, “Replica placement
in data grid: considering utility and risk,” in Proceedings of the
IEEE International Conference on Coding and Computing, pp.
Figure 37: The alarm’s effect. 354–359, April 2005.
22 International Journal of Distributed Sensor Networks

[15] Q. Q. Shi and H. Huo, “By using the steepest descent algorithm [32] L. Tian and Z. T. Liu, “The piecewise linear fitting based on least
to improve the node location accuracy of maximum likelihood squares,” Chinese Journal of Computer Science A, vol. 39, no. 6,
estimation,” Chinese Journal of Application Research of Comput- pp. 482–484, 2012.
ers, vol. 25, no. 7, pp. 2038–2040, 2008. [33] N. A. Gumerov and A. Zandifar, “Structure of applicable
[16] X. B. Wang, M. Y. Fu, and H. Zhang, “Target tracking in wireless surfaces from single views,” in Proceedings of the 8th European
sensor networks based on the combination of KF and MLE Conference on Computer Vision (ECCV ’04), pp. 482–496, 2004.
using distance measurements,” IEEE Transactions on Mobile [34] Y. Yan and A. M. Zhang, “Research on the detection method of
Computing, vol. 11, no. 4, pp. 567–576, 2012. interpolation model of moving target image,” Chinese Journal of
[17] N. Bulusu, J. Heidemann, and D. Estrin, “GPS-less low-cost Computer Knowledge and Technology, vol. 3, no. 18, pp. 22–25,
outdoor localization for very small devices,” IEEE Personal 2007.
Communications, vol. 7, no. 5, pp. 28–34, 2000. [35] M. H. Zhu and H. Q. Zhang, “Research on indoor location
[18] H. Cheng and H. Wang, “Research on centroid localization technology of RSSI-based in wireless network,” Chinese Journal
algorithm that uses modified Weight in WSN,” in Proceedings of Modern Electronic Technology, vol. 17, pp. 12–14, 2010.
of the International Conference on Network Computing and [36] B. Chu and M. Wu, “Saving communication based on gener-
Information Security (NCIS ’11), pp. 287–291, May 2011. alized inverse of non-negative matrix factorization in wireless
[19] J. Wang and H. Jin, “Improvement on APIT localization sensor networks,” Chinese Journal of Zhong Nan University
algorithms for wireless sensor networks,” in Proceedings of Journal, vol. 44, no. 4, pp. 8–12, 2013.
the International Conference on Networks Security, Wireless
Communications and Trusted Computing (NSWCTC ’09), pp.
719–723, April 2009.
[20] K. Yedavalli and B. Krishnamachari, “Sequence-based localiza-
tion in wireless sensor networks,” IEEE Transactions on Mobile
Computing, vol. 7, no. 1, pp. 81–94, 2008.
[21] Z. H. Liu, J. X. Chen, and X. K. Chen, “Research on the
new algorithm of sequence localization in Wireless Sensor
Networks,” Electronic Press, vol. 38, no. 7, pp. 1552–1556, 2010.
[22] S. Joshi and S. Boyd, “Sensor selection via convex optimization,”
IEEE Transactions on Signal Processing, vol. 57, no. 2, pp. 451–
462, 2009.
[23] D. Niculescu and B. Nath, “Ad hoc positioning system (APS),” in
Proceedings of the IEEE Global Telecommunications Conference
(GLOBECOM ’01), pp. 2926–2931, San Antonio, Tex, USA,
November 2001.
[24] J. Zhang and Y. H. Wu, “The location algorithm based on DV-
Hop in wireless sensor networks,” Chinese Journal of Application
Research of Computers, vol. 30, no. 2, pp. 323–326, 2010.
[25] F. Fang, S. Zhao, and P. Guo, “The range analysis of RSSI-based,”
Chinese Hournal of Sensing Technology Journal, vol. 20, no. 11,
pp. 2526–2530, 2007.
[26] T. S. Rappaport, Wireless Communications: Principles and Prac-
tice, Prentice Hall PTR, Upper Saddle River, NJ, USA, 1996.
[27] M. Zhang, S. Zhang, and J. Cao, “Probability-based clustering
and its application to WLAN location estimation,” Journal of
Shanghai Jiaotong University (Science), vol. 13, no. 5, pp. 547–
552, 2008.
[28] A. M. Ladd, K. E. Bekris, A. Rudys, G. Marceau, L. E. Kavraki,
and D. S. Wallach, “Robotics-based location sensing using wire-
less Ethernet,” in Proceedings of The 8th Annual International
Conference on Mobile Computing and Networking, pp. 227–
238, Association for Computing Machinery, Atlanta, Ga, USA,
September 2002.
[29] K. Kaemarungsi and P. Krishnamurthy, “Properties of indoor
received signal strength for WLAN location fingerprinting,”
in Proceedings of the 1st Annual Int Conference on Mobile
Computing and Ubiquitous Systems: Networking and Services,
pp. 14–23, IEEE Press, Boston, Mass, USA, 2004.
[30] J. Sheng and S. Q. Xie, The Theory of Probability and Mathemat-
ical Statistics, Higher Education Press, Beijing, China, 2008.
[31] X. Wang, J. Zhang, and Y. Zhang, “Registration of remote
sensing images based on Gaussian fitting,” in Proceedings of the
3rd IEEE Conference on Industrial Electronics and Applications,
ICIEA 2008, pp. 378–381, June 2008.