2. Advances in
mechatronics have resulted in
widespread of robotics.
Today robots are controlled by computers.
Technical advances are gradually increasing
their similarity to human.
Engineers are attempting to add sensors to
enhance their ability to perform function.
4. Robots could
not perform difficult tasks.
AI to robots gives the power to make decision
and logical inferences.
Robots use sensors and cameras to obtain
information from environment.
Use of microprocessor in robotics is to
enhance their intelligence.
5. ANN
is a network of processing elements
capable of doing processing in parallel and in
distributed manner.
ANN is a mechanism work same as that of
human brain.
ANNS are mathematical models of infoProcessing.
It has two types:
a) Single layer Neural Network
b) Multi layer Neural Network
6. Back
propagation is a technique with
backward error propagation.
It reduces the error from hidden layers.
In back propagation network the activation
function chosen is the tan sigmoid function.
The sigmoid function accommodate large
signals without saturation.
Sigmoid Transfer function is…
Yi = 1/1+e^-X
7. Such
like AI vision is also essential for robots.
Machine vision can be defined as the
acquisition of image, followed by processing
and interpretation of image by computer.
8. A mini
robot was developed which was
capable of pushing the components.
This robot consists of two cameras for vision.
These cameras capture the top and front image
of minimum resolution 128 128.
There is a gripper used to place the
components into the bin.
The gripper and cameras are controlled using
computers.
9. The
captured image should be processed to
obtain the best results.
There are three steps of image processing:
a) Image acquisition
b) Image processing I
c) Image processing II
10.
In this stage the captured image is read by dividing
the bit map image into pixel.
Each pixel has a value that is proportional to the light
intensive of that small proportion of the scene.
The new divided image is stored from bottom to top.
11.
ANN is developed in such a way that it will operate
only on binary codes, i.e. 0 and 1
The captured image contains red, blue and green
ranges.
Colored image should convert to black and white
image.
This image contains a large amount of noise. Various
parameters are used to remove these noises.
After removing the noises the tilted image is obtained.
12.
The size of captured image is varying due to some eternal
factors.
To reset the size of image a technique, Set Partitioning In
Hierarchical Tree (SPIHT) method is used.
By SPIHT normal size image can be obtained.
13. Camera
does not require a driver to operate.
Brightness, resolution and other settings can
change.
Separate module is provided for controlling
the robot.
ANN is generalized network.
Image identification time is faster.
14. Learning time
increases as the number of
training cases increases.
It requires some memory for Database
operation.
If the position of cameras changed then it
requires training .
15. The
developed ANN system has been tested
and the results are satisfactory.
The proposed system can be implemented
without substantial modification, using
available equipment, thus the cost is very low.
This research has two main directions,
a) Improvement in the integration of the system
b) Identification of components at various light
sources and environments.
Notes de l'éditeur
Mechatronics is a design process that includes a combination of mechanical engineering, electrical engineering, control engineering and computer engineering. Mechatronics is a multidisciplinary field of engineering.