Universidad Autnoma de Occidente.

Araque Andres, Salazar Daniel,

Pedroza Francisco. Kobuki Deep learning robot.

PROGRESS 1 OF THE FINAL PROJECT

KOBUKI DEEP LEARNING ROBOT.
Andrs Felipe Araque Mora, [email protected]; Francisco Javier Pedroza Fernndez,
[email protected], Daniel Salazar, [email protected]

AbstractThis first draft of the final project of a lot of robot technology to do

includes the history of the Kobuki robot and some dangerous work.
possible applications with the robot, also includes
all the dimensions of the robot with blueprints and In 2003 was selected as the national
the 3D model made using SolidWorks, the final
excellent manufacturing technology
importation to ROS using an URDF text file and
finally the roslaunch of the launch file to have the center.
Rviz running.
They have been developing a lot of
Index Terms: URDF, ROS, CAD 3D, Kobuki. projects for educational robots, and some
robots for task as cleaning or industrial
robots.
OBJETIVES
The Kobuki robot was built for advanced
To measure the Kobuki robot which is in research in Robotics and Artificial
the UAO laboratory to know all the Intelligence (Deep Learning).
dimensions.
Kobuki is one of the best mobile base
To create the 3D model from the 2D draw designed for education and research.
with the dimensions. Kobuki provides power supplies for
external computer power as well as
To import the 3D model to ROS using
additional sensors and actuators.
URFD file created from scratch.
The kobuki structure have a 3D Depth
Camera, a mobile processor, speaker &
HISTORY microphone.

The TurtleBot2 (Kobuki) was develop by To build the robot applications you can use
Yujin Robotics (Korea) in 2013. It is a the robot operating system (ROS), that
mobile robot of differential kinematics bring us some libraries and more
developed for research and education. characteristic to do it.

The TurtleBot was introduced in April of

2011 as a low-cost personal robot kit
running open-source software.
Since 1998 Yujin Robotics developed
sensors and automation system for some
factories and began the commercialization
Universidad Autnoma de Occidente. Araque Andres, Salazar Daniel,
Pedroza Francisco. Kobuki Deep learning robot.

Deep Learning represents a more intimate

focus to the way the human nervous
system works.

APPLICATIONS
AAL: Ambient Assisted Living
Robots able to provide services in
domestic environment for example, bring
Figure 1. Kobuki base.
medicine, monitor the health status, fall
detection, provides information in real
time in case of some accident using a
human robot interface(HRI).
Navigation, localization and
mapping based on Kinect (Slam
2D y 3D).
Visual odometry algorithms, 3D
algorithms based in Kinect, Localization
of Radiofrequency sources with
directional antennas, etc.

Educational
Educational tool to teach programming
Figure 2. Kobuki mobile robot. and mechatronics.
Education in advanced robotics,
What is deep learning? autonomous behaviors, artificial vision,
localization and mapping, trajectory
Deep Learning is the Approximation to planning, kinematics studies, etc.
Human Perception, logical structures are
used that are more closely resembling the Kobuki can be used for investigation
organization of the nervous system of platforms, depends of the robotic kit used
mammals, having layers of process units in it structure.
(artificial neurons) that specialize in
detecting certain characteristics existing in
the perceived objects. Artificial vision is MORPHOLOGY
one of the areas where Deep Learning A mobile robot of differential kinematics,
provides a considerable improvement with two wheels each wheel gives a half of
compared to more traditional algorithms. the final speed and two dead wheels used
as a support.
Universidad Autnoma de Occidente. Araque Andres, Salazar Daniel,
Pedroza Francisco. Kobuki Deep learning robot.

It has two plates where is possible to attach First the base, on frontal view its possible
a cam or others accessories depending of to see the width of the entire base and the
the task to do, that plates dont have joints width of the part of the base which is
so only apports to the robot the changes on lower.
its center of mass, same with the columns
Also, the different heights including the
that supports the plates.
total one and some lateral heights.
All the measures are in millimeters so the
GEOMETRY total width of the base is something like 35
cm, and the total height is about 11 cm.

Figure 3. Kobuki available in UAO.

Figure 5. Bottom view of the base.
Its important to know the distances
We created the blueprints of the Kobuki between the wheels and the diameter of
getting measures from the UAO robot and each one, because the base will support the
then with that measures we created the 3D wheels.
Model.
The total distance between them is 18 cm
and the distance between a wheel and the
center of the base is 8.8 cm.

Figure 4. Front view of the base.

Universidad Autnoma de Occidente. Araque Andres, Salazar Daniel,
Pedroza Francisco. Kobuki Deep learning robot.

Figure 6. Right view of the base.

The base has a structure where the ports
are located that structure is 27 cm length.
Figure 8. Superior view of the plate.
Then is the plate, that plate supports the
camera and other things user can put on the
robot.
Has a diameter of 31 cm and four little
holes separated between them by 23 cm
and 16 cm.
Also, has a cylinder hole with inner radius
0.73 cm.

PROCEDURE
The first step was measure the robot which
is on the UAO laboratory, then we created
the blueprints of each link, then we created
the 3D solids using SolidWorks.
SolidWorks exports the solid as a STL to
will be exporting as STL to the URDF file.

Figure 7. Front view of the column.

Following the structure the next part are
the columns, that columns are 41 cm
length and has a diameter of 0.68 cm, its
divided on three little structures.
Universidad Autnoma de Occidente. Araque Andres, Salazar Daniel,
Pedroza Francisco. Kobuki Deep learning robot.

Figure 11. URDF robots code.

Figure 9. 3D model in solidworks.

we export the 3D model robot structure as
a base of the mobile robot, the robot base Figure 12. Rviz model.2
is a fixed structure, and the two wheels was
created in the URDF file as cylinder in the
Geometry tag with the real sizes (radius
and length), each wheel have a continuous
type joint specifying the relation of parent
and child link (base and wheel), and the
coordinates of each wheels seen since of
the bases reference frame origin
coordinates.

Figure 13. Rviz model with the wheels

frames.

CONCLUSIONS
Importing a STL model to ROS is
Figure 10. URDF robots code. the best way to have an exact
geometry and the possibility of
make the model functional.
Universidad Autnoma de Occidente. Araque Andres, Salazar Daniel,
Pedroza Francisco. Kobuki Deep learning robot.

Kobuki robot as a differential Wiki.ros.org. (2017). kobuki -

mobile robot offers a lot of ROS Wiki. [online] Available at:
possibilities in terms of possible http://wiki.ros.org/kobuki
applications, from health monitor [Accessed 25 Mar. 2017].
to a creator of maps using its
camera. Wiki.ros.org. (2017).
urdf/Tutorials/Building a Visual
Creating a URDF model from Robot Model with URDF from
scratch help us to understand how Scratch - ROS Wiki. [online]
ROS interprets the entire robot as Available at:
joints and links. http://wiki.ros.org/urdf/Tutorials/
Building%20a%20Visual%20Rob
ot%20Model%20with%20URDF
A good way to put each link on %20from%20Scratch [Accessed
place is using the coordinates from 25 Mar. 2017].
<origin> on the URDF file,
because sometimes is very hard to
put the model on origin at the 3D
model Software.
BIBLIOGRAPHY
yujinrobot. (2017). Home.
[online] Available at:
http://en.yujinrobot.com/
[Accessed 28 Mar. 2017].

Robotnik. (2017). Applications of

the new Kobuki robot TurtleBot2
- Robotnik. [online] Available at:
http://www.robotnik.eu/aplicacion
es-del-nuevo-robot-kobuki-
turtlebot-ii/ [Accessed 27 Mar.
2017].

Ros.org. (2017). POLYRO (oPen

sOurce friendLY RObot) - ROS
robotics news. [online] Available
at:
http://www.ros.org/news/2011/05/
polyro-open-source-friendly-
robot.html [Accessed 29 Mar.
2017].