Multi-Sensor Based Collision Avoidance Algorithm for Mobile Robot

Abstract

Collision Avoidance (CA) systems have been used in wide range of different robotics areas and had extraordinary success in minimizing the risk of collisions. It is a critical requirement in building mobile robot systems where they all featured some kind of obstacle detection techniques in order to avoid two or more objects from colliding. The purpose of this paper is to present an algorithm for performing collision avoidance in mobile robot that is relying on the use of low-cost ultrasonic with infrared sensors, and involving some other modules, so that it can be easily used in real-time robotic applications. The proposed algorithm is implemented in multiple scenarios with several obstacles placed in different locations around the robot. Our experimental run shows that the robot has been successfully detecting obstacles and avoiding collisions along its path.

Full text

Multi-Sensor Based Collision Avoidance Algorithm
for Mobile Robot
Abrar M. Alajlan, Marwah M. Almasri, Khaled M. Elleithy
Computer Science and Engineering Department
University of Bridgeport
Bridgeport, CT
aalajlan@my.bridgeport.edu, maalmasr@my.bridgeport.edu, elleithy@my.bridgeport.edu
Abstract— Collision Avoidance (CA) systems have been used in
wide range of different robotics areas and had extraordinary
success in minimizing the risk of collisions. It is a critical
requirement in building mobile robot systems where they all
featured some kind of obstacle detection techniques in order to
avoid two or more objects from colliding. The purpose of this
paper is to present an algorithm for performing collision
avoidance in mobile robot that is relying on the use of low-cost
ultrasonic with infrared sensors, and involving some other
modules, so that it can be easily used in real-time robotic
applications. The proposed algorithm is implemented in multiple
scenarios with several obstacles placed in different locations
around the robot. Our experimental run shows that the robot has
been successfully detecting obstacles and avoiding collisions along
its path.
Keywords- Collision avoidance algorithm, robotics control , obstacle
detection, mobile robot
I. INTRODUCTION
Collision Avoidance (CA) systems have been used in
wide range of different robotics areas and had extraordinary
success in minimizing the risk of collisions. Collision
avoidance systems are mostly applied in transportation
systems such as aircraft traffic control, autonomous cars and
underwater vehicles etc. Collision avoidance is a critical
requirement in building mobile robot systems where they all
featured some kind of obstacle detection techniques in order to
avoid two or more objects from colliding (two mobile robots
or one robot with an obstacle).
The main purpose of obstacle avoidance is to obtain a
collision-free trajectory from the starting point to the target in
monitoring environments. Obstacles can be divided into two
types, static where the obstacle is predefined and has a fixed
position and dynamic obstacle where the position is not pre-
known and has uncertain motion patterns (moving objects).
Detecting dynamic obstacles is more demanding than
detecting static obstacles since the dynamic has a changeable
direction and requires a prediction of dynamic obstacle
position at every time step in order to achieve the requirement
of a time-critical trajectory planning [1].
The dynamic obstacle algorithm is much more
complex, since it does not only detect an object, but also
consider some kind of measurements regarding the
dimensions of the moving obstacle and distance between the
mobile robot and object. As soon as these measurements have
been determined, the obstacle detection algorithm needs to
steer the robot around the object and continue toward the
target [2]. However, this continuous procedures need to be
performed while the robot is moving toward the original target
[3].
In order to operate efficient collision avoidance
technique, many successful mobile robot systems depend on
the sensing capabilities and collision detection modules of the
robot to obtain collision-free motion [4]. The collision
position can be detected through some measurement coming
from different modules such as cameras and GPS integrated
with and variety of sensors like ultrasonic sensors and Infra-
red sensors [5]. Based on fusion of these readings, the robot
itself can make a decision to detect and prevent collisions
while it is moving along the path, and thus mobility is not
delayed until the detection process is updated [5].
In this brief, a fairly general algorithm is developed
that has components of formation development, obstacles
detection. Contributions of this paper is relying on the use of
low-cost ultrasonic with infrared sensors, and involving some
other modules, so that it can be easily used in real-time robotic
applications.
The paper is organized as follows. Section 2
discusses the literature of collision avoidance techniques.
Section 3 shows the hardware platform for the proposed
approach. Section 4 presents the proposed algorithm of
collision avoidance in robotic systems based on multiple
sensing modules. The algorithm is then implemented and the
overall system design and operation are discussed in section 5.
The experimental results that demonstrate obstacle detection is
presented in section 6. Lastly, section7 offers conclusions.
II. RELATED WORK
Many collision avoidance algorithms have been proposed in
the literature of robotics motion planning that allow the robot
to reach its target without colliding with any obstacles that
may exist in its path. Each algorithm differs in the way of
avoiding static/dynamic obstacles. The concept of Artificial
Potential Field (APF) is recently reported by M. Zohaib et al.
in [6]. APF is a collision avoidance algorithm that always
finds the shortest path from source to destination. It simply
avoids an obstacle by generating a repulsive force from

obstacle to repel the robot and attractive force from target to
attract the robot. Then, the total force on the robot is the sum
of the attractive and the repulsive force. That is, the force is
influenced by how far the robot is from obstacles and target.
When the robot is close to target, its speed will be slow and
vice versa[6]. Even though, APF has a very simple technique
for collision avoidance, it is very sensitive to local minima in
case of a symmetric environment [7].
The Vector Field Histogram (VFH) is a real-time obstacle
avoidance algorithm that used to detect obstacles and avoid
collisions while moving the robot toward its destination [2].
This method uses two-dimensional Cartesian histogram grid,
where it divides the whole area around the robot into small
sectors. In VFH, there is two-stage data reduction process in
order to compute the desired commands. In the first stage, the
two-dimensional histogram is converted to one-dimensional
polar histogram. In the second stage, the algorithm selects the
most suitable sector with low polar density and calculates the
steering angle in that direction [8].However, this algorithm is
not suitable for detecting dynamic (moving) obstacles since its
only studies the distance between the vehicle and obstacles
while ignores obstacles velocities[9].
Sezer et al. proposed a new collision avoidance technique
called Follow the Gap method (FGM). FGM avoids obstacles
by finding the gap between them. Then, it calculates the gap
angel and compares it with the threshold gap. Thus, the robot
will follow the gap as long as it is greater than the threshold
gap [10]. FGM performs three main computations. First, it
calculates the gap array in order to find the maximum gap.
Second, it calculates the center angle of the maximum gap to
ensure safe path from obstacles center. Third, it calculates the
final heading angle by combining both gap center angle with
goal angle. Then, based on these calculations, the robot moves
along final heading angle in order to avoid obstacles [11].
Several other obstacle avoidance algorithms are suitable for
real-time application and will not be discussed here due to
space limitations. Among the reported algorithms, the
proposed one is intended to develop a collision avoidance
mobile robot that is able to detect obstacles and avoid
collisions. The mobile robot is equipped with multiple sensors
and a microcontroller. These multiple sensors are used to get
information about the surrounding area and then make a
decision based on the sensors reading.
III. HARDWARE PLATFORM
.NET Micro Framework is an open source platform
for embedded devices. Using Visual Studio and C#,
developers can create embedded applications [12]. In
addition, Microsoft has recently introduced the .NET
Gadgeteer which is an open-source platform that enables
using .NET Micro Framework and Visual Studio for
combing the benefits of object-oriented programming and
the assembly of small electronic devices [13]. .NET
Gadgeteer is a standardized way to connect mainboards and
modules. One of the most well-known companies that adapt
the use of .NET Gadgeteer and .NET Micro Framework is
GHI Electronics. GHI Electronics offers a variety of
mainboards, sensor modules, and power modules [14].
FEZ Cerbot is a wheeled robot from GHI Electronics.
It has a168Mhz CPU, 1MB FLASH, 2 gears/motors, 16
configurable LEDs, 2 reflective sensors (infrared sensors),
buzzer, and a four AA battery holder. The motors controls
the speed and the direction up to 40V 3A. Also, FEZ Cerbot
has six Gadgeteer sockets that can be used for serial
modules, SPI modules, I2C modules, and I/O modules. FEZ
Cerbot includes one USB Client cable for connecting to PC
[14]. FEZ Cerbot robot is shown in Fig. 1.
Fig. 1, FEZ Cerbot robot
Fig. 2, Reflective sensors on FEZ Cerbot.
The reflective sensor in FEZ Cerbot is an infrared sensor
that is used for edge detection. It detects the presence of
objects in a range. The sensor value changes when an object is
detected within the range. The reflective sensor is shown in
Fig. 2.
Distance US3 Module is used to measure the distance
from an object. It is an ultrasonic sensor that measures the
time between sending a sound wave and the sound echo back
from that object and then calculates the distance. The range
for this module is between 2cm and 400cm [14]. Distance
US3 Module is presented in Fig. 3.

Furthermore, in order to capture images, a camera module is
used. It is an USB Camera that streams images with a
resolution of 320x240 [14]. The camera module is depicted in
Fig. 4.
Fig. 3, Distance US3 Module.
Fig. 4, Camera Module
GPS module is used to determine the current location
coordinates and measures the distance. This module consists
of a U-Blox Neo-6M GPS module and an antenna [14]. Fig. 5
shows the GPS module.
Fig. 5, GPS Module.
IV. THE PROPOSED APPROACH
This section describes the proposed approach for collision
avoidance based on different types of sensor modules. Based
on the output of these modules, the robot will avoid obstacles
by detecting them and change its direction accordingly. The
sensor modules used are infrared sensor for edge section,
distance module which is an ultrasonic sensor that is used to
measure the distance between the robot and the detected
object, camera module that is used to detect objects and
measure the object dimensions and the distance, and the GPS
modules to determine the coordinates and the distance
between two points. Four algorithms are proposed in this
paper. Three algorithms handle the sensor readings where the
fourth algorithm is the main algorithm that incorporates the
other algorithms.
1. Edge detection and distance measurement
In order to detect the edges, the reflective sensors (infrared
sensors) on the robot are used. Two infrared sensors: one on
the left and one on the right. After getting these sensors’
readings, the readings will be compared with a threshold value
within the range of detection. If both infrared sensors values
are less than the threshold, then an obstacle is detected.
Otherwise, use the Distance US3 Module to measure the
distance from an object within 2cm to 40 cm range.
GetDistanceInCentimeters() function is used to convert the
senor readings into centimeters. The detailed algorithm is
shown in Algorithm 1.
2. Camera- based Object detection Algorithm
Camera module is used to capture images. These images are
processed to extract and detect the objects from the image.
After detecting objects, the image will be converted to
grayscale image. After that, thresholding is performed in order
to isolate the object from the background followed by noise
elimination. After extracting the object from the image and
isolate it from the back ground, the object’s dimension will be
measured. By knowing the coordinates of the object in 2-D
environment, then the distance between the robot and the
object is obtained [15]. Algorithm. 2 summaries the Camera-
based Object detection algorithm.
A. Programming Languages
Algorithm..1: Edge detection and distance measurement
Algorithm 1: Edge detection and distance
measurement
In
p
ut: L: Left infrared sensor, R: Right infrared
sensor,
a: ultrasonic sensor reading
Output: β: Edge detection,
gd: distance in cm
begin
set L
set R
// threshold=10
if ((L< threshold) &&(R< threshold))
then set β== True
else
if (a== set value)
// Convert reading to cm
then set gd= GetDistanceInCentimeters()
return gd
end if
end if
end else
end

Algorithm.2 : Camera- based Object detection
3. GPS-based distance measurement Algorithm
By using GPS module, the robot location is determined.
After taking the readings of the robot and the object, a formula
is used to convert degree to radians and later to centimeters
using Haversine formula. Knowing the earth radius, the
distance between the robot and the object is calculated.
Algorithm. 3, explains the process in more details.
Algorithm. 3: GPS-based distance measurement
4. Collision avoidance algorithm
This algorithm is the main algorithm as it incorporates the
other algorithms’ output as an input. The distance between the
robot and the object is measured by the Distance module. The
reflective sensor is used for edge detection. The camera
module is used to detect objects from the image and then
measure the object’s dimensions and the distance between the
robot and the object as well. GPS module is used to measure
the distance between the robot and the object. After obtaining
all these data, sensor data fusion is done and the collision
avoidance decision is made. If the robot detects an object in
the sensing range, it will stop, wait for a second, back up, wait
for another second and then turn. Finally the sensor moves
forward. Algorithm.4 describes the process in details. The
flowchart of the algorithm is shown in Fig. 6.
Algorithm. 4: Collision avoidance algorithm
V. MOBILE ROBOT EXPERIMENTAL SETUP
The collision avoidance algorithm was tested on FEZ
Cerbot robot from GHI Electronics Company. This robot has
Algorithm 2: Camera-based Object detection
Input: I:Image from Camera, gr: convert RGB to
Grayscale, thr: thresholding,
noise: Eliminate all noises, di: Measure object
dimensions
Output: , Δd: distance between object and camera
begin
set I
// using Image processing
Extract obj
// convert color image to grayscale image
Set gr
// for separating the object from the background
Set thr
//Eliminate all noises from the image
Set noise
//Measure the dimensions of obj
Set di
// (x1, x2) coordinates of obj1, (y1, y2) coordinates
of obj2
Set Δd=√((Δx)
2
+(Δy)
2
) =√((x
2
-x
1
)
2
+(y
2
-y
1
)
2
)
return Δd
end
Algorithm 4: Collision avoidance algorithm
Input: s: Motor speed, gd: distance in cm by
ultrasonic sensor, Δd: distance between object
and camera, d: Distance between two positions
by GPS
Output: Collision avoidance
begin
Set s
While (true)
do Edge detection and distance
measurement algorithm
if (β!= True)
then do Camera-based Object
detection
do GPS-based distance
measurement
Dist=Fusion (gd, Δd, d)
// Collision avoidance Decision
if (Dist >= 5&&Dist<= 30)
then set motorSpeed(0,0) // stop
set Delay =1000 ms
set motorSpeed(-s,-s) // Backup
set Delay =1000 ms
set motorSpeed(s,-s) // Turn
set motorSpeed(s,s) //
Foreword
else
set motorSpeed(s,s) //Move Foreword
end if
end else
end while
end
Algorithm 3: GPS-based distance measurement
Input: pos1: position1, pos2: position2
Output: d: Distance between two positions
begin
set pos1 and pos 2
// convert degree to radians
Dlat1=pos 1.lat *PI/180
Dlong1=pos 1.long *PI/180
Dlat2=pos 2.lat *PI/180
Dlong2=pos 2.long *PI/180
Δlat=dlat2-dlat1
Δlong=dlong2-dlong1
// convert from radians to cm using Haversine formula
a=sin
2
(Δ lat)+ cos (dlat1)*cos (dlat2)* sin
2
( Δlong/2)
c=2*a tan (2*(√a,√(1-a) ))
d=R*c // R is the earth radius
return d
end

End
Dist= Fusion
(gd,Δd, d)
If (Dist >=
5 &&Dist
<= 30
)
Yes
No
Move Forward
Stop, Delay, Ba ck
up, Delay, then
Turn
Convert distance
into cm
Camera- based
Object detection
GPS-based di stance
measurement
gd: distance
in cm
Δd:distance
obj &
d: Distance
by GPS
Read from
Distance
Image from
Camera
Read
from GPS
Start
L&R infrared
sensor
If( L< T hr
&& R<Thr
Yes
No
been equipped with multiple sensors as follow. Two infrared
sensors at the front (left and right), ultrasonic sensor (Distance
US3 module), camera module, and GPS module as presented
in Fig. 7. The algorithm were implemented on a Windows
platform using .NET. The parameters used in this experiment
are defined in Table I.
Table. I: parameters used in the experiment
Fig. 6: Flowchart of collision avoidance algorithm
Fig. 7, the mobile robot used in the experiment
Fig. 8, the snapshots of the experimental run
VI. EXPERIMENTAL RESULTS
The main goal of this work is to detect objects within the
range of detection. We deployed multiple sensors in order to
enhance the overall performance of the robot by improving its
reliability and robustness. The most accurate measurement
among them is obtained by performing sensor data fusion.
After data fusion is done, the robot will be avoiding the
obstacles and turn away from any detected obstacles. The
snapshots of an experimental run are shown in Fig. 8.
As shown in Fig. 8, the robot starts sensing the environment
by different types of sensors and moves forward due to no
presence of objects in its way. Once the object is detected, the
distance is computed between the robot and the object based
on these sensor modules. Depending on the calculations, the
robot will stop and then back up. After that, it will turn to
avoid the object on its path and finally move forward. As
depicted in Fig. 8 (3&4), the robot detects another object on
its way and again it avoids it and then turns. This experiment
was implemented in multiple scenarios with several obstacles
placed in different locations around the robot. Our
experimental results show that the robot has been successfully
detecting obstacles and avoiding collisions along its path.
Name Value
Motor speed 70
Sensor Range 2 cm – 40 cm
Threshold for edge
detection 10

VII. CONCLUSIONS
An obstacle avoidance methodology using multiple sensors
for the mobile robot was presented and has been developed
and tested on our experimental mobile robot FEZ Cerbot. The
proposed framework is a real time and portable for any
environment with various shapes of obstacles it encounters
during navigation. The framework was successfully tested
with various shapes of obstacles. The distance fusion
calculated using the collision avoidance algorithm shows the
improved performance in detecting objects and avoiding
collision.
REFERENCES
[1] L. Zeng and G. M. Bone. “Mobile Robot Collision Avoidance in
Human Environments” International Journal of Advanced
Robotic Systems, 2013, Vol.10 doi: 10.5772/54933.
[2] J. Borenstein and Y. Koren. The vector field histogram – fast
obstacle avoidance for mobile robots. IEEE Journal of Robotics
and Automation, 7(3):278–288, 1991
[3] P. Brunn. Robot collision avoidance. The Industrial
Robot 23(1), pp.27-33.1996.Available:
http://search.proquest.com/docview/217012908?accountid=264
84.
[4] D. Fox, W. Burgard, S. Thrun, A hybrid collision avoidance
method for mobile robots, Proceedings of the IEEE
International Conference on Robotics and Automation, 1998,
Vol.2, pp.1238-1243
[5] B.Marco, B.Gabriele, M. Caccia, L. Lapierre. “A Collision
Avoidance Algorithm Based on the Virtual Target Approach
for Cooperative Unmanned Surface Vehicles” Control and
Automation (MED), 2014 22nd Mediterranean Conference of
IEEE, June 2014 pp. 746 – 751
[6] M. Zohaib, M. Pasha, R. A. Riaz, N.Javaid, M. Ilahi, R. D.
Khan. “Control Strategies for Mobile Robot With Obstacle
Avoidance”, Journal of basic and Applied Scientific Research,
Vol. 3 no.4, 2013 pp. 1027-1036.
[7] J.Oroko, B.Ikua,“Obstacle Avoidance and Path Planning
Schemes for Autonomous Navigation of a Mobile Robot: A
Review”. Sustainable Research and Innovation
Proceedings,Vol. 4, 2012
[8] Z. Yan, Y. Zhao, S. Hou, H. Zhang, and Y. Zheng, “Obstacle
Avoidance for Unmanned Undersea Vehicle in Unknown
Unstructured Environment,” Mathematical Problems in
Engineering, vol. 2013, doi:10.1155/2013/841376
[9] Y.Zhu, T.Zhang, J.Song, X.Li, M.Nakamura “A New method
for mobile robots to avoid collision with moving obstacle”
Artificial Life and Robotics,Feb.2012, vol.16(4), pp.507-510
[10] M.Zohaib, S.Mustafa, N.Javaid, A.Salaam, J.Iqbal. “ An
Improved Algorithm for Collision Avoidance in Environments
Having U and H Shaped Obstacles”, Studies in Informatics and
Control, vol.23(1), 2014 pp.97-106
[11] V.Sezer, M.Gokasan,” A novel obstacle avoidance algorithm:
“Follow the Gap Method”. Robotics and Autonomous
Systems”, Journal Robotics and Autonomous System,
vol.60(9), Sep.2012
[12] Netmf.com, '.NET Micro Framework', 2014. [Online].
Available: http://www.netmf.com/. [Accessed: 30- Jan- 2015].
[13] CodePlex, 'Microsoft .NET Gadgeteer', 2006. [Online].
Available: http://gadgeteer.codeplex.com/. [Accessed: 30- Jan-
2015].
[14] Ghielectronics.com, 'Home - GHI Electronics', 2003. [Online].
Available: https://www.ghielectronics.com/. [Accessed: 30-
Jan- 2015]. [15] F. Shahdib, M. Bhuiyan, M. Hasan and H.
Mahmud, “Obstacle Detection and Object Size Measurement
for Autonomous Mobile Robot using Sensor”, International
Journal of Computer Applications, vol. 66, no. 9, 2013.