See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228720781

Fabrication and Control of 4-DOF, Autonomous Robotic Arm Using Low Cost
AVR Controller

Article · January 2005

CITATIONS READS

3 1,400

5 authors, including:

Javaid Iqbal Nida Hassan

National University of Sciences and Technology COMSATS University Islamabad
48 PUBLICATIONS 481 CITATIONS 1 PUBLICATION 3 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Sophia Bano on 04 February 2015.

The user has requested enhancement of the downloaded file.

 Fabrication and Control of 4-DOF, Autonomous Robotic Arm
Using Low Cost AVR Controller

Javaid Iqbal, Nida Hassan, Atif Waqar, Sophia Bano, Rashid Ilyas
Department of Mechatronics, College of E&ME, National University of
Sciences and Technology Rawalpindi, Pakistan,
[email protected], [email protected], [email protected],
[email protected], [email protected]

ABSTRACT

This paper summarizes the design and fabrication of a four degree of freedom
autonomous robotic arm using Digital image processing. The robotic arm boasts a high
performance, RISC architecture based microcontroller manufactured by Atmel, the
AtMega16. Control of the robotic arm has been achieved successfully using four servo
motors, each of which incorporates high resolution quadrature encoders. The robotic
arm is equipped with a pair of high resolution digital cameras. These cameras transmit
real time video to a Pentium based processor. The image processing module searches the
video stream for predefined templates and calculates the coordinates of the object to
grasp and transmits them serially to the microcontroller. The microcontroller implements
inverse kinematics algorithms and then actuates the respective motors through angles so
as to reach the coordinates desired in the most efficient manner. The motors are
actuated using the internal PWM channels of the microcontroller and power amplifiers
using efficient Mosfets. The robotic arm also has the provision of being controlled
without the cameras relying on user input for the starting and terminating points. These
coordinates are fed to the controller directly using a keypad and LCD module
eliminating the Pentium processor. Thus even without a vision system the robotic arm
may be defined as being autonomous. The end effector is a two finger gripper. The
robotic arm has a load bearing capacity of upto 500gms and a maximum workspace of
about 650mm of a meter.

1. Introduction

As manpower becomes more expensive and lives of people more precious, the need to find
cheaper, faster and safer ways of performing tasks grows exponentially. This has led to research
and growth in the field of robotics. The growing need in factory automation for an expanding
involvement of robotic workstations in industrial applications increases the demand for not just
efficient but also light weight robotic arms [1]. However in this era of fast growing technology
there is an imminent need for the manipulator to be cost effective. Many ways of controlling the
arm remotely have been developed, using the microcomputers [2, 3]. However in order for robots
to become more useful, they must be able to function in more demanding environments than
typically found in most factories. Thus if the system is autonomous, it is both time saving and
cuts down on the labor cost.
Computer vision [4] is a fundamental requirement for autonomous machines. Methods of
acquiring information using computer vision vary according to requirements. A pair of audio

DARH2005 Conference, Session 5.3

 Fabrication and Control of 4 DOF Autonomous Robotic Arm p. 2

video cameras has been used to acquire stereoscopic video and the code has been developed on
assumption of uniform background using edge detection [5] and pair of cameras.
The robotic arm developed is a prototype, a light weight and low powered version of an
industrial robotic arm. The goal is to implement all the control concepts and prove them using
this scaled down version. A servo system has been used for the control of robotic arm [6, 7].The
robotic arm is powered by four DC servo motors. The entire control of the robotic arm has been
achieved using one RISC based microcontroller, the Atmel AtMega16 [8]. This controller is not
capable of image processing hence the need of a Pentium based processor aroused.
To achieve precise, accurate and stable system, PI control has been implemented on each
motor [9]. The control terms must be individually adjusted or “tuned” for optimal performance. A
simple PI algorithm has been developed so as to reduce the response time of the system. PI
controller provides a robust closed loop control capable of implementing position control.
A pair of audio video cameras acquire real time video stream and an algorithm searches the
video for predefined objects to calculate their location relative to reference axes. Once the
position coordinates have been calculated, joint angles for each respective motor are determined
using the algorithm based on the inverse kinematics [10].
At89C52 PC

Inverse Digital AV
Kinematics Image Cameras
Processin

ATMega 16 H-Bridge

+ + Power
Angles to Proportional Counts
Encoder Amplifiers Motors
Controller to
Counts
- +
PWM/PW

Integral
Controller

Encoder
Counts

Figure 1 A Block Diagram of the System

2. The Mechanical Structure

The mechanical structure has been designed with the aim of demonstrating the concept of
implementing digital image processing, inverse kinematics and closed loop controls. With this
purpose in mind, the manipulator has been limited to four degrees of freedom. Being a concept
demonstrator, the weight handling capability and workspace has been limited. Aluminum has
been used due to it light weight, ease of machinability and low cost.
All the four joints are revolute, three of which form a planar manipulator and the fourth is
the waist for the planar arm. The global workspace is the set of points (x, y, which lie in the
vertical plane and Φ3 is the angle of the end effector with the horizontal) that can be reached by
the end effector for all specified orientations Φ3 of the output link r3 (where r3 is the length of the
last link).
The manipulator consists of 3 links of lengths 380mm, 250mm and 210mm respectively
and a turn table having a diameter of 500mm. It features a two finger hybrid gripper as the end
effector. The gripper can grasp regular shaped objects like cylinders with convenience. Four 24V,
3200 rpm servo motors have been used to drive each of the four joints. The reduction ratio of the
gearboxes for the planar robotic arm is 130:1. The speed of the waist has been reduced by a ratio

DARH2005 Conference, Session 5.3

 Fabrication and Control of 4 DOF Autonomous Robotic Arm p. 3

of 500:1. The entire mechanical structure has been modeled in ProE and a stress analysis of every
individual component has been carried out using Ansys 9.0.
Limit switches are also mounted on the links. Their purpose is to ensure that no physical
motion goes out of range. Thus limit switches here act as override controls for the arm segments.

Figure 2 The 4 DOF Manipulator Figure 3 Camera Used

3. Vision System

Computer vision is a fundamental requirement for autonomous machines. Methods of acquiring

information using computer vision depend upon the specific application. Typical techniques use
camera calibration for machine vision. In contrast, our technique doesn’t require calibration of
cameras. Object detection is implemented using edge detection and a pair of cameras assuming
uniform background. For the calculations we use distance between optical centers of cameras, the
angles which the cameras make with a predefined line and field of view [11].
For range and offset, horizontal field of view is utilized. Law of sine is used in
calculation.

Figure 4 Principle Arrangement of Cameras

In figure 4,
β=ζ+ι
γ=χ+κ
where, ζ and χ are angles which each camera makes with the line BC and ι, κ are
angles calculated from image. Solving above triangle give us range and offset of target from our
desired camera.
For height calculation, actual height of camera lens is measured and is not disturbed
thereafter. The perpendicular imaginary line is drawn from camera to reference height and its
corresponding pixel (cp) is found. Now utilizing triangulation, angle for target point is found. The

DARH2005 Conference, Session 5.3

 Fabrication and Control of 4 DOF Autonomous Robotic Arm p. 4

side opposite to target angle is calculated. If pixel corresponding to point of target is less than cp
then the side calculated is added to referenced height and vice versa.

4. Kinematics

To reach the desired position and orientation motors, the joint angles are needed. These joint
angles are calculated using the inverse kinematics and the motors are then actuated accordingly.
An inverse kinematics code developed runs from the microcontroller At89C52 calculates
the joint angles for the respective motors. The code has been optimally written as the controller
has only 256 bytes of on chip RAM. The algorithm is based on the numerical approach. Due to
the mechanical limitations, only elbow up solution is used. The joint angles are governed by the
following equations:

Figure 5 Kinematic Model

x = l1cosθ1 + l2cos(θ1+θ2) + l3cos(θ1+θ2+θ3) (3)

y = l1sinθ1 + l2sin(θ1+θ2) + l3sin(θ1+θ2+θ3) (4)
Φ = θ1 + θ2 + θ3 (5)
There is a second mode of operation also. This provides the user with a manual override,
a hex keypad and LCD has been interfaced with At89C52. Instead of using the coordinates from
the vision module, the controller prompts the user for the start and end point of the end effector.
These parameters are then used to calculate the joint angles. These angles are then serially
transmitted to the AtMega16 for implementing closed loop controls.

5. Controller

Atmel AtMega16 is a high performance, low power, 8 bit microcontroller. It features an on chip
in system programmable flash memory of 16 KB and 1024 bytes of SRAM, 4 PWM channels all
of which are used for motor control and 3 external interrupts which are used for reading the
encoders. The controller has two modes of serial communication, the standard USART and the
SPI protocol for master slave configuration. AtMega16 has a maximum throughput of 16 MIPS.
Based on the RISC architecture, most of the commands are executed on single clock cycles. The
combination of these features makes the controller suitable for control applications.
The joint angles from the inverse kinematics module are received by the AtMega16. Each
of the motors is then rotated by their respective angles. This is done by implementing closed loop
controls on each of the motor. The closed loop controls are implemented using the PI
compensators.

DARH2005 Conference, Session 5.3

 Fabrication and Control of 4 DOF Autonomous Robotic Arm p. 5

5.1 PI Compensator

Each of the motors used has built in quadrature encoders. The encoders generate square
waves at a frequency proportional to the angular velocity of the motor shaft. There is a phase
difference of 90o between the waveforms of each of the two encoders.
The joint angles calculated by the inverse kinematics algorithm are translated into the encoder
counts using the following equation.
r (t) = (Фd / 360) * (ec * gr) (6)
Where r (t) =total counts
Фd=desired angle
ec=encoder counts per revolution
gr=gear ratio
The number of counts thus calculated is treated as the error. At time t=0 the error is
maximum and is equal to the calculated counts. The proportional part of the controller ensures
that the motor gets an output voltage proportional to the error. Greater the error, the greater is the
applied voltage. As the motor rotates in the desired direction the error starts decreasing and hence
the output voltage to the motors. During this motion the encoders generate the pulses which are
counted. This is the real time position of the shaft and is represented by y (t).
yp (t) = KP*{r(t) – y(t)} (7)
Where Kp is the proportional gain
Since under loaded conditions the error can never be zero, the integral controller is
implemented to eliminate this steady state error. The integral error is defined as:
And the output of the integral controller is
yi(t) = Ki*{ e(t) + e(t-1)} (8)
The two expressions for the output are superposed to get a function that eliminates the
error most efficiently.
The new output function is
y(t) = Kp* e (t) + Ki*{ e(t) + e(t-1)} (9)

6. Results

The design, fabrication and control of an autonomous robotic arm is a challenging task. The
interfacing of a vision system makes the task more complicated. Results showed that the light-
weight mechanical arm performed well when a PI controller is used. The efficient use of an
isotropic material, such as aluminum, results in improved rigidity. The reduction of the link
masses enables actuators to accelerate the joints very quickly. The image processing algorithm
developed is fast enough to generate the coordinates of the object so that the time response of the
system is not compromised. The errors in the calculated coordinates are within specified
tolerances. The graph below shows the error trend in the calculated depth and the actual depth.
0.5
0.4
0.3
0.2
Error (cm)

0.1
0 `
-0.1 0 20 40 60 80 100
-0.2
-0.3
-0.4
Actual Distance (cm)
Figure 5 Error Trend in Range

DARH2005 Conference, Session 5.3

 Fabrication and Control of 4 DOF Autonomous Robotic Arm p. 6

Error can be reduced by increasing resolution, by decreasing viewing angle of cameras and by
increasing distance between optical centers of cameras. The inverse kinematics code is efficient
so as to ensure that the RAM of the microcontroller is not exceeded; the results have been
verified by both algebraic and geometric methods. The PWM channels and the interrupts of
AtMega16 made the implementation of PI control easy. The entire system performed as desired,
the accuracy of the robotic arm was less than 0.5cm along each axis.

7. Conclusions
The control of light weight 4 DOF autonomous robotic arm has been described. To improve
performance, the arm is made of aluminum. The coordinates of the object are generated by an
algorithm using a vision system based on a pair of audio video cameras. These coordinates are
then used to calculate the joint angles using an inverse kinematics code embedded in
microcontroller At89C52. Position control has been implemented on these angles using a PI
controller through the microcontroller AtMega16.

References

[1] Michael P. Hennessey, Joseph A. Priebe, Paul C. Huang and Richard J. Grommes, “Design of a light
weight robotic arm and controller”, Proc of IEEE Work on Robotics Systems, pp. 779- 784, 1987.
[2] John N. Lygouras, Basil G. Mertzios and Nicholas C. Voulgaris, Design and construction of a
microcomputer-controlled light-weight robot arm, Robotics and Autonomous Systems 7, 269-283,
1991.
[3] N. Ahuja, U. S. Banerjee, V. K. Darbhe, T. N. Mapara, A. D. Matkar, R. K. Nirmal, S. Balagopalan,
th
“Computer Controlled Robotic Arm”, Proceedings of the 16 IEEE Symposium on Computer-Based
Medical Systems (CBMS’03), 2003.
[4] Mana Saedan and Marcelo H. Ang Jr.,’3d Vision Based Control on An Industrial Robot”, Iasted
International Conference On Robotics And Applications, Nov 19-22, 2001, Florida, USA.
[5] Digital Image Processing By Rafeel C. Gonzalez, Richard E. Woods, 2nd Edition ,Pearson,2002
[6] B.P Lathi, “Signals, Systems and Controls”. Harper and Row, New York, 1974.
[7] B.C Kuo, “Automatic Control Systems”, Prentice-Hall, Eaglewood Cliffs, N.J .1982.4th Edition
[8] www.Atmel.com last accessed on March 2005
[9] www.jashaw.com last accessed on January 2005
[10] John. J. Craig, “Introduction to Robotics”, Mechanics and Control, 2nd Edition, Addison-Wesley,
1989.
[11] www.dr-lex.34sp.com/qtvr/viewangle.html

DARH2005 Conference, Session 5.3

View publication stats