TELKOMNIKA
Vol. 10, No. 4 612
ceiling, it is possible to reduc be used by different robots s
RFID solutions for instance.
3.1. Overall Architecture
The proposed system resorts to a camera with exter
a unique and distinctive color The synchronization
infrared signal modulated in fr SCL present in the environm
signal from the robot and trigg the infrared signal back to the
detected and the landmark i Figure 1. This way the landma
out approach. Since the envi created on the ceiling and t
accuracy required by the robo
Figure 1. System architecture
3.2. Localization procedure
The localization algo should be performed before th
the current position of landma instance, using a conventiona
that must be adjusted based o limit to the number of LEDs
environment increases or wh combination of LEDs in eac
identification.
The second step focu implements the SincroVision
height that will be further com image, defines the navigation
possible to obtain the required
After this, during the the range defined in relation
distance are considered more the LEDs that are closer to t
LEDs means that the LED bel
4, December 2012 : 609 – 620 uce the obtrusive level of the original environment.
simultaneously as they do not experience any oc
em is based on two components: the Robot and ternal trigger for the synchronization with the SCL,
sequence. n between image acquisition and the SCL is acc
frequency. This signal is sent by the robots trigge nment. Whenever a landmark is in range, it rece
iggers their visual RGB LEDs. At the same time, the he robot and the camera shutter is triggered. The
k is identified by the algorithm based on a priori mark is able to respond to several robots using a F
nvironment is populated by several SCL, an artifici d the distance between each landmark is direct
bot to perform its tasks.
ure: The robots localization module and the SCL la are placed on the ceiling.
gorithm presented in this paper is based on five the localization process since it retrieves the prio
arks in the environment. This means that each SC nal EKF-SLAM approach. Each SCL consists of a
d on the size of the environment where the robots m s used by each SCL and they should increase a
when higher localization accuracy is required. Ob each SCL must be unique in order to allow a
ocuses on cluster detection, which is a pre-proc concept. Briefly, this step calibrates the camera
omputed based on the distance between the LED ion axes of the robot, and finally, the color calibrati
ed color palette. cluster filter stage all clusters of pixels whose
on to a typical size of a LED are eliminated. The ore reliable and, therefore, the localization phase
the robot a large distance between one LED a elongs to another cluster.
ISSN: 1693-6930 nt. This system can
occlusion such as
nd SCL. The robot , which must have
ccomplished by an ger controller to the
ceives the infrared the landmark sends
e cluster of LEDs is iori knowledge, see
FIFO first-in, first- ficial constellation is
ectly related to the
landmarks that
ive steps. The first priori knowledge on
SCL is identified, for a number of LEDs
s move. There is no as the size of the
bviously, the color a more complete
ocessing stage and era, the landmarks
Ds captured in the tion. This way, it is
se size is not within he SCL at shorter
e will only consider and a set of other
TELKOMNIKA
Indoor Localization Syst Furthermore, each clu
Identification step. Since the possible to identify the LEDs
position of each LED at the g possible to estimate the robo
every landmark that will be u color combination of detected
the position of each LED in searching all previously impor
The following and fina one SCL described in the glob
navigation reference, it is pos 2depicts the localization conc
L2 LED. It is important to hi
value of the LED i and descr The first step to comp
global
,
and the naviga Θ
ோ
X
ୖ
X
ଵ
X
ଵ ଵ
cos Θ
ୖ ோ
ଵ
ଵ
ଵ
sin Θ
ோ
Figure 2. Localization base Given the virtual line
and L2 and computing their same angle but described in th
the robot Θ
ோ
. Equations 1, define the minimum number o
position is used in order to com ISSN: 1693-6930
ystem based on Artificial Landmarks and …. Andry cluster set of LEDs with closer distances is ident
he SCL can be distinguished by their unique color Ds based on the a priori knowledge and, theref
global coordinate reference belonging to a detect bot’s pose. Thus, the LEDs Identification stage a
used in the localization step. The identification is ed LED sand; therefore, it will be possible to provid
in the world reference frame. The identification orted artificial landmarks.
inal stage is the localization stage. Given the pos lobal reference and the position of the same LEDs
ossible to determine the position and orientation of ncept based on a cluster formed by one green L1
highlight that
is the value for the Cartesian com cribed in the frame j, and vice-versa for
. pute the robots pose
ோ
,
ோ
, Θ
ோ
is finding the a igation frame
ଵ
,
ଵ
, that is, the robots orientatio
ୖ
Y
ଵ ଵ
sin Θ
ୖ ோ
ଵ ଵ
cos Θ
ோ
ased on a single landmark formed by 2 LEDs green e dashed blue, see Figure 2, which passes by
ir slope - the angle described in the global refe n the frame of the robot, it is possible to determine
, 2 and 3 are used to compute the localization r of LEDs in each cluster SCL as two. In these
ompute the robots position, however, it is also pos
dry Maykol G. Pinto 613
ntified in the LEDs or combination, it is
refore, knowing the cted landmark; it is
aims at identifying is achieved by the
vide information on n is performed by
ositions of LEDs of Ds described in the
of the robot. Figure L1 and one yellow
omponent X, is the angle between the
ation. 1
2 3
een and yellow. the two LEDs L1
ference and - the ne the orientation of
ion of the robot and equations, the L1
ossible to use L2.
TELKOMNIKA
Vol. 10, No. 4 614
The algorithm presen step a priori knowledge, wh
architecture. 4. Results and Discussion
Several experiments the localization system. The t
circumstances and different en 4.1. Robustness
The first step to analy may compromise the integrity
brightness on the accuracy of compare two cases: the detec
lighting. The results below p resulting images of the pre-p
were previously defined are r necessary to identify in the ori
To reduce the influen CCD camera is intentionally m
enters the camera this is why
Figure 3a. Original camer external illuminat
Figure 3c. Pre-processed illumination the original came
This happens becau
4, December 2012 : 609 – 620 ented is cyclically executed over time, with the exc
hich can be performed only at the beginning or e
were conducted in order to evaluate the performa e tests were made over a large periods of time, di
environment conditions. alyze the performance is confirming robustness ag
rity of the visual detection. The impact and influ of the localization system were tested. This way,
tection of the SCL with and without the presence o provide the image obtained by the camera Fig
processing stage Figures 3b and 3c, where on represented, that is, pixels that have colors of th
original image. ence of the environment light sources, the apertur
minimized. This way it is possible to reduce the am hy Figure 3a is darker.
era image with ation.
Figure 3b. Pre-processed im identification of the image
illumination Fig.
sed image with color identification of the image with mera image was not presented because it is quite s
ause every the environment light sources were turn ISSN: 1693-6930
xception of the first even in the SLAM
mance achieved by different navigation
against factors that fluence of ambient
y, it was possible to e of intense artificial
Figure 3a and the only the colors that
f the LEDs that are
ture of the onboard amount of light that
image with color ge with external
ig. 3a.
ithout external similar to this one.
urned off.
TELKOMNIKA
Indoor Localization Syst These Figures show
each LED, with and without ex directly in front of the environm
which means thatthe localizat was no problem related to an
the typical size of a LED. neglected.
Figure 4a. Original camer external illumination. Similar
LED and the environment LEDs with a color iden
for the system see Figures4 completely detected since the
verify consistency such as LE the captured information in ord
In these experiments an external illumination which
illuminations to power up each environments since the tests d
4.2. Error characterization