=Paper=
{{Paper
|id=Vol-2498/short59
|storemode=property
|title=Indoor magnetic positioning with an array of magnetic field sensors
|pdfUrl=https://ceur-ws.org/Vol-2498/short59.pdf
|volume=Vol-2498
|authors=Byungmun Kang,Jiseung Jung,DaeEun Kim
|dblpUrl=https://dblp.org/rec/conf/ipin/KangJK19
}}
==Indoor magnetic positioning with an array of magnetic field sensors==
https://ceur-ws.org/Vol-2498/short59.pdf
Indoor Magnetic Positioning with An Array of
Magnetic Field Sensors
Byungmun Kang1 , Jiseung Jung2 , and DaeEun Kim∗
1
Biological Cybernetics Lab, Yonsei University,Seoul, Korea
2
Biological Cybernetics Lab, Yonsei University,Seoul, Korea
3
Biological Cybernetics Lab, Yonsei University,Seoul, Korea
http://cog.yonsei.ac.kr, {kbmang,roadrunner23,daeeun}@yonsei.ac.kr
Abstract. One of the most advanced navigation technologies in recent
years is called SLAM. It is a technology that maps the environment
through a robot’s sensing system while exploring an unknown environ-
ment. This allows the robot to position itself and to reach its final des-
tination through the created map.
In this paper, we introduce a base technology to apply to SLAM using
our new magnetic field sensing system. A magnetic field sensing sys-
tems were applied instead of visual sensors or distance sensors for the
localization of robots in an indoor environment instead of a visual and
distance sensors in this paper. Using surrounding magnetic field system
can estimate the current global position of the robot and the direction
of nearby magnetic objects. Finally, we show correct positioning of the
robot through experiments using odometrical and magnetic field infor-
mation in a real world learned indoor environment.
Keywords: Magnetic Field Positioning, Navigation, Magnetic Field Sens-
ing System.
1 Introduction
There has been a considerable amount of development in navigation in recent
years. Among the technologies that have emerged in this area, SLAM is one
of the key technologies for autonomous driving. SLAM can identify the current
location and position relative to the overall environment to estimate the desti-
nation [1–3]. The technology underlying this SLAM technology is positioning.
Positioning allows the robot to determine its position, speed and path. It is also
possible to obtain errors of its position on a given map and update its posi-
tion. For example, path integration technology allows location information to be
known using only internal elements such as the rotation of the robot’s motor or
wheel size as the robot moves [4, 5]. However, odometrical errors tend to accumu-
late due to mechanical errors of the motor or slipping on the floor surface as the
actual robot moves. To overcome this, there are many studies that use various
sensing systems. For instance, robots use visual sensors or recently Lidar sensors
to obtain location information and the distance between surrounding objects and
2 Byungmun Kang, DaeEun Kim and Jiseung Jung
(a) (b) (c)
Fig. 1. Magnetic Field Sensing System (a) The system is attached to the Mobile
Robot(Turtlebot). (b) A diagram of the array of magnetic field sensors.
the robot. Thus, before the navigation technology is implemented, positioning
and localization technologies identify location and status information.
In this paper, we introduce a magnetic positioning technology using a magnetic
field sensing system. This magnetic field technology is also based on biomimetic
technology. As we all know, there is a huge magnetic field signal in nature com-
ing from the Earth and many animals and insects use this geomagnetic field for
survival. The ultimate goal of this study is to obtain the positioning technology
based on magnetic field.
To do this, we applied the magnetic field sensing system to the positioning tech-
nology in an experimental environment and updated the robot’s position using
magnetic field information. We performed real-life experiments by attaching a
magnetic field sensing system made by our team to a mobile robot. The exper-
iment was conducted in a compact space inside a laboratory. We measured the
sensor values in the sensing system and conducted mapping based on magnetic
objects. We used a total of 72 points at 10 cm intervals and checked the limi-
tations and the sensitivity of the sensing system. Through this study, we were
able to verify the possibility for laying the foundation of magnetic field based
navigation technology.
2 Method & Experiments
2.1 Magnetic Field Sensing System
The magnetic field sensing system we introduced in this paper consists of eight
analog sensors and a 12-V DC solenoid by placing eight of these analog magnetic
field sensors in a circular 45 degree interval, as shown in Fig. 1. From this mag-
netic field sensing system, we can obtain two kinds of signals. We define these
signals as DC and AC signals as shown in Fig. 2. The DC signal is defined as
the change in voltage that can be obtained when the solenoid is switched on.
The AC signal refers to the amplitude of the voltage change when the solenoid
is switched on. Details of the configuration of this sensing system are described
in [6].
Indoor Magnetic Positioning with An Array of Magnetic Field Sensors 3
2
Normal state Normal state
Metal Sphere 0.2 Metal Sphere
1.8
0.1
voltage(V)
voltage(V)
1.6
DC signal 0
1.4
-0.1
1.2
-0.2
AC Signal = Difference of and
1 -0.3
0 2 4 6 8 10 0 2 4 6 8
time(s) time(s)
(a) (b)
Fig. 2. The two type of signals the sensing system can detect. (a) is DC signal, (b) is
AC signal.
2.2 Experimental Setting
The experiment was carried out in a indoor laboratory of 350cm x 400cm size in
Fig. 3. Three computer desktops and monitors were placed on the outside, and
two metal trash cans were placed in the center. The surrounding magnetic field
of experimental environment was measured by the magnetic field sensing system
of the robot at each point as it moved at intervals of 10 cm through a total
of 72 points. Using the data obtained from these experimental environments,
a Magnetic Experience Map (to be introduced in the following sections) was
constructed. Also, as shown in (d) of the Fig. 3, seven steps were performed
and the current location was updated through our matching method with the
magnetic experience map.
ƵƚƚŝŶŐ
'HVNWRS
ϵ͕ϴ ϯWƌŝŶƚĞƌ DŽŶŝƚŽƌ ĞƐŬƚŽƉ
DĂĐŚŝŶĞ
'HVNWRS
7UDVK
&DQ
0RQLWRU
0RQLWRU
dƌĂƐŚ dƌĂƐŚ
ĂŶ ĂŶ
^ƚĂƌƚ
'HVNWRS
7UDVK
&DQ
3ULQWHU
ŶĚ
'
0DFKLQH
&XWWLQJ
0DFKLQH
ƵƚƚŝŶŐ
&XWWLQJ
Ϯ͕ϭ
ϭϬĐŵ
DĂĐŚŝŶĞ ĞƐŬƚŽƉ DŽŶŝƚŽƌ ĞƐŬƚŽƉ
ϭ͕Ϯ ϭ͕ϭ
ϭϬĐŵ
(a) (b) (c) (d)
Fig. 3. Experimental Environment. (a) is the front view, (b) is the side view and (c)
is the diagram of experimental environment. (d) is the path of the robot test based on
user’s command.
2.3 Creating the Magnetic Experience Map
There are two main ways to form the magnetic experience map : using the
largest of eight sensor values and then expressing all eight sensors in Fig. 5.
The first method shows how a magnetic field is topologically formed for the
4 Byungmun Kang, DaeEun Kim and Jiseung Jung
0DJQHWLF'LUHFWLRQ
6HQVRU
6HQVRU 6HQVRU
6HQVRU ࡿ
6HQVRU
6HQVRU 6HQVRU
ࡿ
ࡿ
6HQVRU
ࡿ ࡿ ࡿ ࡿ ࡿ ࡿ ࡿ ࡿૠ
ࡿ
ࡿ
ࡿ
ࡿ
ࡿ
(a) (b)
Fig. 4. The method of creating a magnetic map and estimating the magnetic direction.
experimental environment. The second method allows us to visually estimate the
direction of the magnetic objects. The map is created according to the two types
of signals mentioned earlier, DC and AC. Through this map, we implemented
robot behavior using existing memory or experience.
2.4 Magnetic Direction
As mentioned in the system configuration section earlier, this magnetic field
sensing system uses eight sensors for sensing the surrounding magnetic objects
as shown in (b) of Fig. 4. Since the sensing system is circular, it can be used to
detect all directions as the robot moves. The magnetic field sensing information
can be used to determine or estimate the direction of the surrounding magnetic
object. This method was motivated by sand scorpions and was implemented
through the following formula.
m
zeiφ = zk eiφk = x + yi (1)
k=1
m
m
x= zk × cos(φk ), y= zk × sin(φk ) (2)
k=1 k=1
φ = arctan(y/x) (3)
φ is the direction of the vibration signals, and the score for each leg of the sand
scorpion is zk (m=8), and the degree of each leg is φk , which can be expressed
as above equations [7]. In Eq. (1), zk ei φ can be expressed with cosine and sine
values through Euler formula. Then, according to the score and degree of each
leg, it can be expressed as Eq. (2). The values of x and y can be obtained through
Eq. (2) and the direction of the vibration’s source (φ) can be estimated through
Eq. (3) [7, 8]. Therefore, the direction of the vibration’s source can be obtained
using the AC and DC signals instead of the arrival time of the vibration signal
with the above equations.
Indoor Magnetic Positioning with An Array of Magnetic Field Sensors 5
(a) (b) (c) (d)
150 40
35
30
100
25
20
15
50
10
5
0 0
(e) (f) (g) (h)
Fig. 5. Data expression of the magnetic map using the maximum value of sensing
values. (a) is the intensity of value sensing value, (b) is value the x-axis, (c) is value
the y-axis and (d) is value the z-axis. Magnetic Mapping(Experience Map). (a) and
(c) is the DC value of the given environment. (b) and (d) is the AC value of the given
environment.
2.5 Magnetic Matching Method
In this paper, we update the robot’s position through matching method with the
magnetic field information measured in the magnetic environment map. This
matching method is largely divided into two stages. The first is to compare
the information obtained from the magnetic direction and the experience map
obtained at a specific location. At this time, comparing directions would result
in several candidate groups. To reduce this candidate group, we compare the
magnetic field sensing values. This is the second step. When the candidate group
that came through these two steps and the current robot is ordered to go to a
specific location, the point closest to that particular location is finally selected.
3 Results
3.1 Global Direction
Through the magnetic field sensing system, we can estimate the global magnetic
direction of the robot. In other magnetic field sensing systems, when detecting
surrounding magnetic objects through one magnetic field sensor, the magnetic
object cannot be utilized as a compass due to disturbance. In contrast, since
the system has multiple sensors, sensors that do not receive disturbance from
magnetic objects can recognize signals from the magnetic field of the earth. As
shown in Fig. 6, signals are obtained by rotating each sensor clockwise with
regard to the east direction. The signals from the eight sensors can be obtained
with specific x,y values according to their direction, and can therefore detect the
global direction of the robot.
6 Byungmun Kang, DaeEun Kim and Jiseung Jung
S0 S1 S2 S3
0.2 0.2 0.2 0.2
x axis x axis x axis x axis
0.15 y axis 0.15 y axis 0.15 y axis 0.15 y axis
0.1 0.1 0.1 0.1
DC value [V]
DC value [V]
DC value [V]
DC value [V]
0.05 0.05 0.05 0.05
0 0 0 0
-0.05 -0.05 -0.05 -0.05
-0.1 -0.1 -0.1 -0.1
-0.15 -0.15 -0.15 -0.15
-0.2 -0.2 -0.2 -0.2
0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360
Degree Degree Degree Degree
(a) (b) (c) (d)
S4 S5 S6 S7
0.2 0.2 0.2 0.2
x axis x axis x axis x axis
0.15 y axis 0.15 y axis 0.15 y axis 0.15 y axis
0.1 0.1 0.1 0.1
DC value [V]
DC value [V]
DC value [V]
DC value [V]
0.05 0.05 0.05 0.05
0 0 0 0
-0.05 -0.05 -0.05 -0.05
-0.1 -0.1 -0.1 -0.1
-0.15 -0.15 -0.15 -0.15
-0.2 -0.2 -0.2 -0.2
0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360
Degree Degree Degree Degree
(e) (f) (g) (h)
Fig. 6. The DC value of each sensor during rotating. For the z-axis, the values are
constant because the magnetic field sensing system rotates in place with the same
height, and data are obtained. (a): sensor 0, (b): sensor 1, (c): sensor 2, (d): sensor 3,
(e): sensor 4, (f): sensor 5, (g): sensor 6, (h): sensor 7.
DC Magnetic Direction AC Magnetic Direction
(a) (b)
Fig. 7. The Magnetic Direction of DC and AC. (a) is DC Magnetic Direction, (b) is
AC Magnetic Direction.
3.2 Magnetic Direction
In this paper, the robot can detect the direction of magnetic objects around it
using information obtained in the experimental environment. We can obtain the
magnetic direction by using the formula (1),(2),(3) in the method section. This
can be applied to avoidance by recognizing magnetic obstacle as the robot moves
in the actual indoor environment, and also to determining the exact location of
its route. As shown the Fig. 7, the direction is given for magnetic fields with
strengths larger than a certain size. This allows for the representation of DC
signals and AC signals, respectively. Where DC signals can be determined at
various points in the macro, AC signals can be determined at a location close to
their own objects, but there is some noise. In fact, in the [6] paper studied by
us, the difference between DC and AC was demonstrated through experiments.
A specific path was set based on the magnetic object (metal trash can), and the
Indoor Magnetic Positioning with An Array of Magnetic Field Sensors 7
200
Without Mag sensor
Error(Euclidean Distance) [pixels]
With a Mag sensors
With 4 Mag sensors
150 With 8 Mag sensors
100
50
0
Pos 1 Pos 2 Pos 3 Pos 4 Pos 5
(a) (b)
Fig. 8. The trajectory of the robot experiment(a) and the comparison of Errors(b).
White line is the user’s command and red dashed line is the trajectory without the
magnetic field information and the blue dashed line is the trajectory the robot with the
magnetic field information. The green dashed is the trajectory using only four sensors’
information and the cyan dashed line is the trajectory using only one sensor placed on
the head of the robot. The yellow circles are the updating positions.
matching method was used to update the current position using magnetic field
information while giving a command to the robot. In fact, in [6], we studied the
effect of DC and AC signals and it was demonstrated through experiments.
3.3 Robot Positioning
The results were compared with the results of moving the robot without mag-
netic field information on a designated path based on user’s command and using
magnetic field information. Also, the results were divided into three types. The
first type is to update the path of the robot using magnetic information from
one magnetic field sensor placed in the head direction of the robot. As shown in
cyan color line of Fig. 8, based on the surrounding magnetic field information
obtained from one sensor, the results are updated with an odometrical error.
Since one sensor is used to find the magnetic direction we performed a rotation
action. In the 5 update positions, the average error is about 48.71 pixels (29.22
cm) which is similar to a method that does not use a magnetic field. Basically,
one pixel is corresponding to 0.6 cm for distance. The second type is to use only
four sensor information out of a total of eight sensors in the front, rear, and side
directions (e.g., north east and west). This type was less rotated than the first
type, and the robot can find the magnetic direction in the magnetic experience
map and update it, as shown in the green dashed line of Fig. 8. The average
error of the second type is 39.75 pixels(23.85cm). The last type is to use all eight
sensor information that can be obtained from magnetic field sensing systems.
Unlike the previous first and second types, the robot can update its position
without rotation action as shown in the blue dashed line of Fig. 8. The average
error is also 24.63 pixels (14.77 cm). The average error value for 5 iterations of
robot test is 47.45 pixels (28.47 cm) when the magnetic field information is not
used as shown in the red dashed line of Fig. 8, and if the surrounding magnetic
8 Byungmun Kang, DaeEun Kim and Jiseung Jung
field information of the magnetic sensing system is used, it can be seen that the
robot can reach the specified position based on user’s command in (b) of Fig. 8.
4 Conclusion
In this paper, by using the magnetic field sensing system developed by us, a
magnetic experience map of the experimental environment could be obtained,
and the global direction could be determined through the array of magnetic field
sensors. In addition, the robot was positioned in an indoor laboratory environ-
ment and the direction of the magnetic object around it was used to position its
current position. Basically, sensing systems and methods for determining the ori-
entation of magnetic objects were constructed based on biomimicry. In the robot
experiment based on the command of actual users, the results were better than
the experiments obtained by using only the robot’s odometrical information, and
we obtained more accurate results by using 8 sensor data of 2 types through the
array of 8 sensors and the solenoid at the center. This proved the sensing system
suitable for the positioning technology based on indoor navigation.
Acknowledgement
This work was supported by the Institute of BioMed-IT, Energy-IT, and Smart-
IT Technology (BEST), a Brain Korea 21 plus program, Yonsei University.
References
1. Wolcott, Ryan W., and Ryan M. Eustice. ”Visual localization within lidar maps for
automated urban driving.” 2014 IEEE/RSJ International Conference on Intelligent
Robots and Systems. IEEE, 2014.
2. McDonald, John, et al. ”6-DOF multi-session visual SLAM using anchor nodes.”
(2011): 69-76.
3. Milford, Michael, and Gordon Wyeth. ”Mapping a suburb with a single camera
using a biologically inspired SLAM system.” IEEE Transactions on Robotics 24.5
(2008): 1038-1053.
4. McNaughton, Bruce L., et al. ”Path integration and the neural basis of the’cognitive
map’.” Nature Reviews Neuroscience 7.8 (2006): 663.
5. Franz, Matthias O., and Hanspeter A. Mallot. ”Biomimetic robot navigation.”
Robotics and autonomous Systems 30.1-2 (2000): 133-153.
6. Kang, Byungmun, and DaeEun Kim. ”Detecting a Sphere Object with an Array of
Magnetic Sensors.” International Conference on Simulation of Adaptive Behavior.
Springer, Cham, 2018.
7. Kim, D. ”Neural network mechanism for the orientation behavior of sand scorpions
towards prey.” IEEE transactions on neural networks 17.4 (2006): 1070-1076.
8. Adams, Samantha V and Wennekers, Thomas and Bugmann, Guido and Denham,
S and Culverhouse, Phil F, ‘Application of arachnid prey localisation theory for
a robot sensorimotor controller”, Neurocomputing, Vol. 74, No. 17, pp. 33353342,
2011.