Design of a Low-cost Autonomous Mobile Robot

K.K. Jinasena & R.G.N. Meegama
International Journal of Robotics and Automation, (IJRA), Volume (2) : Issue (1) : 2011 1
Design of a Low-cost Autonomous Mobile Robot
Kasun K. Jinasena kasun@sjp.ac.lk
Faculty of Applied Science
University of Sri Jayewardenepura
Gangodawila, Nugegoda, Sri Lanka
Ravinda G.N. Meegama rgn@dscs.sjp.ac.lk
Department of Statistics & Computer Science
Faculty of Applied Science
University of Sri Jayewardenepura
Gangodawila, Nugegoda, Sri Lanka
Abstract
Detection of obstacles during navigation of a mobile robot is considered difficult due to varying
nature in terms of size, shape and location of the obstacle as well as ambient light that interferes
with infra-red (IR) signals of the robot. In this paper, we propose a novel low-cost method that
successfully guides a robot along a path using image processing and IR sensor circuits. A high
continuous IR signal is converted into a low continuous IR signal by means of a demodulation
circuit that enables a peripheral interface controller to receive this low continuous IR signal and
take relevant decisions based on the signal. The images taken from a web camera are
preprocessed to remove noise and detect edges. Subsequently, an image processing routine
effectively calculates the angle to be rotated of the front wheels using a scan line algorithm. A
minimum mean distance error of 2.45 was observed in tracking the path at a signal-to-nose ratio
of 26.50. The accuracy of speech recognition was 92% for two voice training sessions.
Keywords: Mobile Robot, Infra-red Signals, Voice Recognition, Robot Navigation, Infra-red
Communication
1. INTRODUCTION
Current research challenges on intelligent autonomous robots include navigation, task planning
and coordination, learning and adaptation, cooperation between robots and humans [1]. Each
model incorporates its own perceptual, modeling, and planning requirements.
Recent studies indicate a growing trend towards tracking of paths using mobile robots. A four
wheel robot powered by a single motor was developed in the University of Aveiro [2] that uses
complex mechanical techniques for both front and rear wheels. By using high-end firewire
cameras and a built-in laptop computer, it managed to navigate along a path interpreting traffic
lights. A unique two-body mobility platform with a hybrid electric power system was introduced in
[3] resulting in an intelligent robotic vehicle with a sensor suite. In subsequent years, a three-
wheeled, differential driven vehicle that uses two brushless smart DC servo motors and a hybrid
system that provides sufficient power for the vehicle to run continuously at a length of more than
10 hours was developed [4]. The four sensors, (digital camera, scanning laser range finder, digital
compass, and differential global positioning system (GPS)), were integrated into a streamlined
system architecture with a laptop serving as the computational core of the system.
All these research finding over the years have contributed immensely to the advancements in
robotics applications. More recent studies in mobile navigation techniques include autonomous
flying vehicle [5], mine detecting vehicles to be used in military applications [6], wall climbing
robots where robotics vehicles have been able to walk along a vertical plane using
electromagnetism, electrostatic adhesion or air suction [7], medical surgery where the motion of

the heart is tracked in three-dimension [13] and micro robots used in biomedical applications to
remove plaque in arteries [14].
Although we are able to come across many applications in robotics that addresses a specific area
of research, one important fact that must be considered when navigating an autonomous robotic
vehicle outdoors is interferences caused by sunlight [15]. Especially when the robot is guided by
IR sensors, there must be a mechanism to distinguish between the sun’s IR rays and the rays
reflected back from an obstacle located at a particular distance away from the moving vehicle.
FIGURE 1: Torque charactisircs of stepper motors
2. METHODOLOGY
This research has its own hypothesis such as the track is on a plane surface with different color
to its background, there are no junctions, the environment is static or dynamic, and the track with
locations of obstacles are not known at the beginning of its navigation. A mechanical structure
was designed for the problem involved with electronic solutions for the speed controllers and
motor drivers. Image processing, speech recognition and text to speech software were used for
the guidance algorithms.
The following sections describe the methodology we have used in designing and implementing
the autonomous mobile robot.
2.1 Chassis Design
The chassis provides a solid support for the electronic boards, camera and sensors. Although
aluminum pipes are the preferred material to build the robot, iron pipes were selected due to their
availability and rigidness. The vehicle moves on three wheels. The two front wheels are each
powered by a stepper motor while the rear wheel, hinged freely to the chassis, turns
automatically.
The platform is made up of iron and was designed with modularity in mind, i.e., the chassis
should provide a solid support for electronic boards, camera and any other items to be mounted
in future.
The two main points of concern in selecting a material for the chassis are rigidness and lightness
of the material as lightness of the chassis helps save power and enhances transmission. On the
other hand, rigidness assists the vehicle to be stronger and carry more weight.

FIGURE 2: Wheel setup of the vehicle
2.2 Wheel Design
Unipolar 24 V, 800 mA per winding (30 ) stepper motors are used to power the front wheels.
Each step motor will rotate 7.5 degrees as there are 48 steps for a single rotation. Each motor
has maximally 0.2 Nm pull-out torque and 0.16 Nm pull-in torque with 100 steps per second with
a maximum of 500 steps per second. A gear system is also used to increase the torque by 3
times. It consist of two wheels having teeth at a ratio of one to three to increased the torque by
three times while reducing the speed by three times.
Fig. 1 illustrates the torque characteristics of the stepper motors while Fig. 2 indicates the setting
up of wheels of the robot vehicle where the front two are steering wheels each powered by a
stepper motor and the rear are free wheels. The vehicle turns by changing speed of its front
wheels while rear wheels turn automatically. This will allow the robot to maintain more stability
than a three wheel vehicle.
The vehicle transmission consists of two stepper motors responsible for steering the vehicle
along the surface. Although rear driven wheels are suitable for heavy vehicles such as buses and
lorries, front drive is used mainly for lighter vehicles so that they can be easilly manoeuvreable.
As such, a front driven mechanism is chosen for our robotic vehicle. Two motors are connected to
the front wheels of robot to make turning easier.
2.3 Obstacle Detection
The vehicle is able to detect certain obstacles along its path and take relevant decisions based on
the type of the obstacle. Ultrasound and Infra Red (IR) are the two most widely used obstacle
detection methodologies in Robotics [8,9]. By using Ultrasound, the velocity and the direction of a
moving obstacle can be identified easily through the Doppler effect. Because ultrasound waves
are susceptible to temperature and humidity, IR, whose speed is constant over a particular
media, is used in most mobile digital electronics [12].
Our vehicle employs IR to detect obstacles. The main challenge in using IR for this purpose is to
distinguish between the robot's and nature’s IR at the receiving end. Under strong sunlight, it is
possible that a high amount of IR signals may incident on the sensor. As such, we need to
overcome the natural IR exerted by the ambient light. One of the solutions to overcome this issue
is to investigate the intensity of the two IR signals as natural IR signals has a lower intensity
compared to the robot's. Natural IR, however, is not constant over a given period. As such, a

modulated (pulsed) IR signal, obtained with frequency generators, is used to overcome this
problem. This modulated pulse signal is shown in Fig. 3.
FIGURE 3: IR signal modulation
As depicted in Fig. 4, the 1
st
modulator converts the continuos signal into a 13 kHz square wave
and then the 2nd
modulator encloses that 13 kHz signal in a 10 Hz signal.
FIGURE 4: Demodulating a reflected signal
Finally, the transmiter emits this signal as IR rays. IR is heat radiation which is absorbed readilly
by black rough surfaces while reflected by white shiny surfaces. Therefore, in order to make the
robot sensitive to all obstacles equally, once the IR module receive signals reflected by an
obstacle with an output of 10kHz (square signal), it is converted to a continuos signal by the
demodulator. The result is then sent to a PIC microcontroller to make necessory decisions.
2.4 IR Signal Demodulation
The task of the Demodulator is to demodulate the signal originating from the IR module and
regenerate a continuos signal. This is achieved by using a capacitor and properly biased
transistors. This transistor circuit causes the signal to be invert but programmatically corrected.
Fig. 5 shows the circuit diagram of the demodulator.
FIGURE 5: IR Signal demodulation circuit

2.5 Stepper Motor Driving Circuit
Fig. 6 shows only a single channel of the stepper motor driver circuit. The actual circuit carries
eight of these circuits for eight windings of two motors. We have integrated a mechanism to
protect the IN 4001. A photo coupler makes the isolation between high voltage and the low
voltage sections. IN4001 protects the D313 from back electro motive force (EMF) of the motor.
FIGURE 6: Stepper motor driving circuit
2.6 Image Processing & Vision
The input video stream is captured via web camera of 1.3 mega pixels resolution and the frames
are extracted as images. These images are then pre-processed to remove Gaussian noise and
detect edges.
As in Fig. 7, two lines scan the acquired image to detect the edges of the path. The angle θ of the
inner and outer covertures is separately calculated. At each scan line, the curvature is calculated
based on the intersection of lines and the edges. Based on the value of the angle θ, the left and
right wheels are stopped to make the turns to left and right, respectively.
FIGURE 7: Detection of path using curvature
At this research, we carried out experiments to determine the vehicle’s ability to track a path that
consists of two edges. These edges have been drawn in black color on a white background.
More complex tracks with multiples edges will obviously require much advanced image
processing routines and fast computing power.
Fig. 8 depicts the main information flow of tracking the edges of a path using image processing
technques. It can be seen that the vechile makes a decision to turn the wheels based on the
location of detected edges.

2.7 Speech Recognition
We know that every speech recognition system consists of an engine that translates sound
waves into text and a list of speech commands. In this research, we used the SAP15 engine that
is integrated into Windows XP operating system. This engine was trained using windows voice
training wizard in which a profound usage of grammar increases the probability of
misinterpretations.
FIGURE 8: Flow chart of path tracking procedure
As such, we tried to keep the set of grammar as small as possible without lose of information.
Grammar, which contains the commands “Go”, “Left”, “Right” and “Stop”, is then fed into an XML
file.
2.8 Main Board
A photographic illustration of the main circuit board is given in Fig. 9 showing the IR sensor and
associated circuitry. The main circuit board constist of 16F877A PIC as shown in Fig. 10. The
pins 37- 40 are being used for the four voice commands given to the vehicle while pins 27 – 30
are reserved for the right stepper motor driver. The left stepper motor is driven by pins 19 – 22.

FIGURE 9: Circuit board of the vehicle.
FIGURE 10: Main circuit board
2.9 PCB Design
Fig. 11 illustrates the printed circuit board (PCB) of the vehicle.
2.10 Mechanicla Structure
Fig. 12 shows the dimensions of the robotc vehile. The net weight of the vehicle was found to be
1.2 Kg that includes all the circuitry, wheels, web cam and IR sensor.
The selected dimensions of the robotic vehicle makes it easier to manipulate along a path with
sharp edges. Also, the compact size of the structures requires less power to drive forward or
backwards.

FIGURE 11: PCB of the robotic vehicle.
FIGURE 12: Dimensions of the robot: as seen from (a) above and (b) right hand side.
3. RESULTS AND DISCUSSION
3.1 Hardware and Software Requirements
A Personal Computer with a Pentium IV (or above) class Processor, 3.0 GHz, 512 MB or above
RAM capacity and 10 GB hard disk space are preferred. All USB ports should be enabled and a
microphone, speakers/head phone and web camera are the other accessories that are required.
We used Microsoft Windows XP operating system with Microsoft’s speech engine 5.1 and DirectX
9 with .NET framework 2.0 and a device driver for the web cam and Useport.sys for port access
are required.
3.2 Transmission
As shown in Table 1, the vehicle is able to travel forward or backward with 3.95 inches per
second on a flat surface while it turns 360 degrees around any of its front wheels, clockwise in
17.35 seconds and anti clock wise in 17.02 seconds. It travels in a straight line with an accuracy
of more than 98.64 %. Speed can be increased by decreasing the period of delay while power
decreases.

Characteristics Result
Forward travel 3.95 inches per second
Forward travel deviation 1.36%
Clockwise rotation 17.36 seconds per 360º
Anti-clockwise rotation 17.02 seconds per 360º
TABLE 1: Transmission characteristics of the robotic vehicle.
3.3 IR Sensitivity
At the beginning of the research, an IR sensor typically found in a computer mouse was used to
detect obstacles. It managed to detect obstacles at a distance of more than one foot with a single
IR light emitting diode (LED). The major problem, however, was sun light. Since this IR sensor
does not have a filter, we found a malfunctioning system even under minor sun light at dawn. The
IR modules that are used in televisions, resulted filtering a pre define frequency from sun’s IR
rays. It was seen that IR modules band pass filter was highly sensitive to 13 kHz IR. Therefore, a
modulator and demodulator circuits were designed to support this frequency.
It was also observed that high sun rise still can affect obstacle detection system because the
frequency of mixed signal is dominated by the frequency of the signal that has the highest
amplitude. This is easily solved by using more IR LEDs that causes the system to be over
sensitive under minor sun light. In a practical situation, such as a remote control system for toys,
the user is allowed to select the sensitivity level manually.
The distance between the IR transmitter and the IR receiver affects the sensitivity as shown in
Fig. 14. Increasing the distance between the IR transmitter and the IR receiver decreases the
sensitivity whereas decreasing the distance between the IR transmitter and the receiver
increases the sensitivity. Results in Table 2 were obtained by changing the distance between the
transmitter and the receiver for a white color, landscaped, A4 size obstacle.
TABLE 2: Distance vs width of sensitivity
3.4 Color Sensitivity
It was evident that the color of the object was a major factor that determines the correct detection
of objects. Fig. 15 shows the sensitivity region of a landscape, A4 size obstacle for the 8 colors.
As shown in Table 3, black had a significant deviation while other colors had minor deviations.
3.5 Voice Recognition
It was witnessed that all voice commands did not have an equal accuracy of recognition. The
accuracy was highly dependent on the speech pattern of the user.
Distance between
transmitted & receiver
(cm)
Straight detection
distance (cm)
Width of detection
region (cm)
3 24 22
4 20 20
5 17 17
6 16 14

Color Distance (cm)
White 24
Yellow 22
Red 20
Orange 19
Green 21
Blue 18
Brown 19
black 3
TABLE 3: Color vs distance of sensitivity
The voice of the speaker needs to be trained first and then, the accuracy can be increased by
having more voice training sessions with the system. Fast speech causes less accuracy or even
unrecognized commands while long phrases resulted in a higher accuracy than shorter phrases.
The accuracy also tends to get low with increasing number of commands. When two or more
commands have similar sound characteristics, the accuracy of those commands decreases. Due
to disturbances from ambient noise, a good, high quality microphone was used. Table 4 gives the
results of accuracy of the system after one voice training session. Each command was tested ten
times with different tones. However, we observed an increase in the accuracy when the number
of training sessions was increased. Table 5 provides the average accuracy when the number of
voice training sessions is two. More training sessions will be required if the vehicle is to recognize
voice commands selected from a large vocabulary.
The speech engine was able to recognize voice of different persons for the same set of voice
commands.
TABLE 4: Average accuracy for one voice training session.
Command Average accuracy (%)
Go 72
Turn left 67
Turn right 68
Reverse 67
Stop 45
Start CV 62
Close CV 66
Who are you 82
Exit 76

Command Average accuracy (%)
Go 72
Turn left 87
Turn right 88
Reverse 87
Stop 65
Start CV 82
Close CV 86
Who are you 92
Exit 66
TABLE 5: Average accuracy for two voice training sessions.
3.6. Path tracking
Direct application of a high pass filter to the captured image resulted in noise at the output that
was highly affected by the path detection module. A significant enhancement was gained by
applying a low pass filter to the captured image prior to applying the high pass filter.
The results obtained from image processing were obviously dependent on the resolution of the
camera used. In order to keep the cost of the vehicle at a minimum, we used a widely used web
camera that can be purchased from any computer store. Also, scanning horizontally on an edge
at the middle of the image resulted in several issues especially when there was an object on its
side. It was avoided by changing the scan order from the middle to up to the edge. A slight
discontinuity or non-uniformity of the road did not make a significant effect to the vision process.
Table 6 provides measured mean distance errors when the vehicle tracks a path under different
signal to noise ratios (SNR). The mean distance error e was calculated as
dsss
l
e
s
s
)'()(min
1
)(
'
vvvv,
v
−= ∫ ∈
where v is the path of the robot’s movement, v is the actual path of the track and l is the length
of v [10].
TABLE 6: Effect of SNR on mean distance error.
SNR Mean distance error
26.50 2.45
22.60 3.35
20.15 4.42
18.75 5.66
17.85 6.57
17.30 13.11
16.50 15.63

3.7 Safety of PIC Microcontrollers
Unlike other electronic equipments, PIC microcontrollers are vulnerable to high voltages. It was
observed that both hardware and software errors resulted in a damaged PIC. It must be noted
that the PIC 16F877A microcontroller is able to provide an output of a maximum current of 200
mA while any of its pins provides an output of a maximum current of 25 mA. Exceeding any of
these limits causes a burnt microcontroller. Interferences from other electrical signals also affect
the inputs to the microcontroller. Most of such problems, however, can be solved by grounding
the pins with a higher resistor, such as 10 K.
3.8 Communication
Currently, the robot does all the communication over wired media. Wireless technologies, such as
Bluetooth or Wi-Fi, can be introduced so that communication can be done from anywhere even
over the Internet. Bluetooth uses the 2.4 GHz radio frequency band to communicate with other
Bluetooth devices [11]. However, the issue that needs to be addressed in this scenario is that a
majority of popular Bluetooth USB dongles are slave devices without any inbuilt firmware. When it
connects to a computer, the computer provides the necessary protocol stack. On the other hand,
if it connects to PIC microcontroller, both the USB protocol stack and the Bluetooth protocol stack
should be implemented inside the PIC. Since a PIC executes only a single program at a time, a
dedicated PIC is needed for this purpose. Although this research uses a PIC16F877A
microcontroller, a PIC 18F series also provides the USB interface. IR sensors
IR Sensors tend to get saturated when the distance between the obstacle and the sensor fall
below 40 cm. Therefore, it is difficult to use it to measure distances less than 40 cm. When
measuring distances that vary between 50 to 70mm, the accuracy of IR sensor ranges from 92 to
95 percent whereas the accuracy of Ultrasound sensor was 90 to 97 percent.
The proposed research show that modulated IR still can be used to detect obstacle even when
they are located close to as much as 3 cm. The speed of Ultrasound waves varies with
temperature and the humidity] and as such needs correction to obtain correct distances. Since IR
is electromagnetic, it neither depends on temperature nor humidity.
4. CONCLUSION
This research presents the design, implementation and results of a low cost minatory model of an
autonomous land vehicle based on computer vision, speech recognition and IR transmission.
The main feature developed by this project was its ability to track a path avoiding disturbances
from sun light using IR, image processing and computer vision. Moreover, the robot can make
voice response to its user. Both the path and location of obstacles were not known in advance.
Results indicate that the robot is able to recognize speech at higher accuracy when the number
of training sessions is increased. We checked the performance of the robot by adding Gaussian
noise along the path to be tracked. With the current setup, it successfully tracked the path when
the signal to noise ratio of the image is higher than 17.30.
Extending the research for the vehicle to transmit data using wireless or Bluetooth technology is
straightforward.
5. REFERENCES
[1] P. Alfonso, J. Azevedo, C. Cardeira, B. Cunha, P. Lima and V. Santos, “Challenges and
solutions in an autonomous driving mobile robot competition”, In Proceedings of the Controlo
Portugese Conference in Automatic Control, Lisbon, 2006.
[2] T. Fong, I. Nourbakhsh, and K. Dautenhahn, “A Survey of Socially Interactive Robots”,
Robotics and Autonomous Systems, 42(3-4):143- 166, 2003.

[3] [online] http://www.avt.me.vt.edu/index.htm.
[4] C.F. Reiholtz, “Virginia Tech Autonomous Vehicle: Jhonny-5”, Department of Mechanical
Engineering, Virginia Tech, USA.
[5] L.O. Mejias, S. Saripallai, P. Cervera and S. Sukhatme, “Visual Surveying of an Autonomous
Helicopter in Urban Areas Using Feature Tracking”, Journal of Field Robotics, 23(3): 185-
199, 2006.
[6] E. Cepolina and M. Zoppi, “Cost-effective Robots for Mine Detection in Thick Vegetation”, In
Proceedings of the 6th
International Conference on Climbing and Walking Robots, Catania,
Italy, 2003.
[7] B. Aksak, M.P. Murphy and M. Sitti, “Gecko Inspired Micro-Fibrillar Adhesives for Wall
Climbing Robots on Micro/Nanoscale Rough Surfaces”, In Proceedings of the IEEE
International Conference on Robotics and Automation, 2008.
[8] G. Benet, F. Blanes, J.E. Simo and P. Perez, “Using infrared sensors for distance
measurement in mobile robots”, Robotics and Autonomous Systems, 40(4): 255-266, 2002.
[9] V.K. Sehgal, R. Sharma, N. Nitin, D.S. Chauhan, A. Srivastava, A. Kumar, Y. Agerwal and
A.M. Khan, “Obstacle Sensing and Anti-falling Sensor Robot Using Embedded Processor”,
In Proceeding of the 11th
International Conference on Computer Modelling and Simulation,
2009.
[10] L.H. Staib and J.S. Duncan,” Model-based deformable surface finding for medical images”,
IEEE Transactions in Medical Imaging, 15(6): 859-870, 1996.
[11] Y.C.F. Amin, S.H.M. Fisal, N.Bakar, “Bluetooth enabled Mobile Robot”, In Proceedings of
the IEEE International Conference on Industrial Technology, 2002.
[12] W. Fehlman and M.K. Hinders, “Mobile Robot Navigation with Intelligent Infrared Image
Interpretation”, Springer, 2010.
[13] R. Richa, P. Poignet and C. Liu, “Three dimensional Motion Tracking for Beating Heart
Surgery Using a Thin-plate Spline Deformable Model”, International Journal of Robotics
Research, 29(2-3): 218-230, 2009.
[14] J.J. Abbot, K.E. Peyer, M.C. Lagomarsino, L. Zhang, L. Dong, I.K Kaliakaktos and B.J.
Nelson, “How Should Microbots Swim?”, International Journal of Robotics Research,
28(11-12): 1434-1447, 2009.
[15] G.Anderson, C. Sheesley , J. Tolson, E. Wilson and E. Tunstel, A Mobile Robot System for
Remote Measurements of Ammonia Vapor in the Atmosphere, In Proceedings of the IEEE
International Conference on Systems, Man and Cybernetics, 2006.

Design of a Low-cost Autonomous Mobile Robot

More Related Content

What's hot (18)

Similar to Design of a Low-cost Autonomous Mobile Robot (20)

More from Waqas Tariq (20)

Recently uploaded (20)

Design of a Low-cost Autonomous Mobile Robot