=Paper=
{{Paper
|id=Vol-2329/paper-04
|storemode=property
|title=Learning Advanced Robotics with TIAGo and its ROS Online Tutorials
|pdfUrl=https://ceur-ws.org/Vol-2329/paper-04.pdf
|volume=Vol-2329
|authors=Jordi Pages,Luca Marchionni,Francesco Ferro
|dblpUrl=https://dblp.org/rec/conf/erf/PagesMF18
}}
==Learning Advanced Robotics with TIAGo and its ROS Online Tutorials==
https://ceur-ws.org/Vol-2329/paper-04.pdf
Learning Advanced Robotics with TIAGo and
its ROS Online Tutorials
Jordi Pages, Luca Marchionni, Francesco Ferro
PAL Robotics S.L.
c/ Pujades 77-79, 4-4, 08005 Barcelona, Spain
jordi.pages@pal-robotics.com, luca.marchionni@pal-robotics.com,
francesco.ferro@pal-robotics.com
Abstract. Learning Robotics in these recent years has become more
easy thanks to ROS and its worldwide success [1]. ROS has managed to
reduce the learning curve for newcomers as results can be obtained even
before going into the theoretical basis like differential steering systems,
forward and inverse kinematics and motion planning. This paper presents
the comprehensive set of on-line tutorials featuring TIAGo, the mobile
manipulator of PAL Robotics, which use Gazebo simulator and ROS to
take a widespread audience to a trip to discover in a practical way a broad
range of areas like control, motion planning, mapping and navigation and
2D/3D perception.
1 Introduction
Dynamic simulations of different robots are very common nowadays. Accurate
simulation models provide an easy way for rapid prototyping and validation of
robotic tasks. Focusing in ROS based simulations, the abstraction layer provided
by frameworks like ros control 1 [2], ensures that what you get in simulation is
what you will get in the real robot, at least in terms of interfaces. Furthermore,
working in simulation is very important when learning robotics, as programming
a real robot may be dangerous for the robot itself, the environment or people
around. Simulation then is an essential tool for learning robotics.
PAL Robotics has published a comprehensive set of on-line tutorials based on
its mobile manipulator TIAGo 2 . The tutorials are organized in different blocks
which are depicted in Figure 1. The first section explains in detail how to get
a computer ready to follow the tutorials presented afterwards. The installation
instructions for an Ubuntu computer and a ROS distribution are explained along
with all the required ROS packages from PAL Robotics in order to have the
dynamic TIAGo’s simulation in Gazebo up and running.
All the tutorials are based on spawning the model of TIAGo in a given
simulated world in order to perform a specific task. The tutorials are structured
in a way that provides the basic ROS instructions in order to have the simulation
1
http://wiki.ros.org/ros control
2
http://tiago.pal-robotics.com
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 30
30
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
2
running, how to run specific ROS nodes performing the target task and also
providing some theoretical background of the robotic task when needed.
The first block of tutorials addresses basic control aspects of the robot. Af-
terwards tutorials about laser-based mapping, localization and autonomous nav-
igation are presented. The next block teaches different ways to perform motion
planning with TIAGo including a pick and place demo. Right after a set of
tutorials showing different computer vision applications are included. Then, tu-
torials about processing point clouds show basic perception tasks based on 3D
data obtained with the depth camera of the robot.
2 Control
These tutorials aim to teach how to command the different Degrees of Freedom
(DoF) of TIAGo.
2.1 Differential Drive Base
The first two tutorials in this block show how to control the differential drive
base of TIAGo in order to have the robot moving in the ground plane. The
main teaching of these tutorials is that a differential drive system can perform
a Twist [3], i.e. the composition of a linear velocity along the axis of the robot
pointing forward and a rotation speed about the center of the robot base.
One of the tutorials instruct on how to send the appropriate message. For
example, the message to have TIAGo moving at 0.5 m/s forward and at the same
time rotating at 0.2 rad/s about its vertical axis can be seen in the following
command line instruction:
rostopic pub /mobile_base_controller/cmd_vel \
geometry_msgs/Twist "linear:
x: 0.5
y: 0.0
z: 0.0
angular:
x: 0.0
y: 0.0
z: 0.2" -r 3
2.2 Upper body joints
The goal of these tutorials is to provide the reader with knowledge about the
typical controllers in ROS based robots to control them in joint space and what
can be achieved with these basic controllers.
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 31
31
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
3
Fig. 1. Overview of the on-line tutorials of TIAGo
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 32
32
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
4
Joint Trajectory Controllers Two tutorials are provided so that the reader
can have a grasp of the different ROS Joint Trajectory Controllers 3 defined in
the upper body of TIAGo each one controlling the following groups of joints:
– arm torso: group composed of the prismatic joint of the torso and the 7
joints of the arm
– gripper or hand: groups of the 2 joints of the gripper or the 3 actuated
joints of the humanoid hand of TIAGo
– head: the 2 joints of the head composing the pan-tilt mechanism
The tutorials provide links to the official ROS documentation explaining how
a Joint Trajectory Controller is useful to execute joint-space trajectories defined
by a set of waypoints so that the trajectory results from interpolation using
different strategies. The use of TIAGo’s joint trajectory controllers is exemplified
with a simple C++ code that the user can easily modify and produce complex
motions with the robot in simulation.
This tutorial can be used when teaching robotics in order to quickly validate
forward kinematics computations of the arm given a joint space configuration by
comparing the pose of the end-effector obtained in simulation. For example, in
order to obtain the cartesian coordinates of TIAGo’s arm tip frame the following
ROS instruction can be used
rosrun tf tf_echo /torso_lift_link /arm_1_link
which will print the transformation from the arm tip link, i.e. arm 7 link, with
respect its parent link, i.e. torso lift link, which is presented in the following form
- Translation: [0.155, 0.014, -0.151]
- Rotation: in Quaternion [0.000, 0.000, 0.020, 1.000]
in RPY (radian) [0.000, -0.000, 0.039]
in RPY (degree) [0.000, -0.000, 2.242]
This is also a good point to introduce the ROS Transform Library 4 to the
students and explain the two rotation representations provided, i.e. the Quater-
nion based and the Roll-Pitch-Yaw representation.
Head control This tutorial provides an example of a more sophisticated con-
troller that can be build on top of the joint-space trajectory controller of the
2 joints of the head. This new controller is the Head Action controller 5 . The
goal of this controller is to have the robot’s head pointing to a given direction,
i.e. looking at an specific cartesian point. Therefore, the input of the controller
is not in joint-space but in cartesian space. The source code of the controller,
3
http://wiki.ros.org/joint trajectory controller
4
http://wiki.ros.org/tf2
5
http://wiki.ros.org/head action
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 33
33
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
5
available in 6 , is also a good example on how to implement inverse kinematics
using the Kinematics and Dynamics Library (KDL) 7 .
The tutorial is based on a simple ROS node implemented in C++ which also
shows how to use the actionlib interface of ROS 8 . The C++ example opens a
window with the RGB image providing from TIAGo’s head camera so that the
user can click on any pixel and the robot in simulation then moves the head so
that it looks at the given direction, see Figure 2.
Fig. 2. Head control of TIAGo defining a sight direction
The C++ example also shows how to make use of the intrinsic parameters
of the camera in order to compute a cartesian point in the line of sight defined
by the pixel clicked by the user.
Playing back pre-defined motions The last tutorial in the Control block
explains how to store in a yaml file upper body motions defined in joint space
so that they can be played back at any time with the Play Motion library in
ROS 9 . The tutorial also explains how to change the velocity of the motions by
varying the time given to reach each waypoint of the trajectory.
3 Autonomous Navigation
Two tutorials are presented in order that lead to autonomous navigation of
TIAGo in simulation. The first tutorial is devoted to explain how to generate
a map with the laser range-finder of the robot. This tutorial can be used as
demonstration of Simultaneous Localization and Mapping (SLAM) by using the
gmapping algorithm wrapped in ROS 10 , see Figure 3.
6
https://github.com/pal-robotics/head action
7
http://www.orocos.org/kdl
8
http://wiki.ros.org/actionlib
9
http://wiki.ros.org/play motion
10
http://wiki.ros.org/gmapping
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 34
34
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
6
Fig. 3. Laser-based map generated by TIAGo
Once the user has created a map a tutorial showing the autonomous navi-
gation of TIAGo is presented. This tutorial embraces theoretical concepts like
Adaptive Monte Carlo Localization [4], particle filtering to track the pose of
the robot in the map, the costmap concept, laser scan matching and obstacle
avoidance with motion planning.
The navigation tutorial also introduces a more sophisticated approach to
obstacle avoidance which consists in using not only the laser scan but also a
virtual scan produced from the point cloud of the depth camera of TIAGo’s
head.
In Figure 4 a global trajectory that TIAGo is trying to follow to reach a
destination point in the map is shown in blue. Furhtermore, the costamps appear
overlayed in the map defining inflated areas that produce repulsion points taken
into account by the local planner that reactively corrects the robot trajectory in
order to avoid obstacles detected by the laser and the camera.
4 Motion Planning
The block of Motion Planning focues on MoveIt! 11 [5], which supports several
motion planning libraries like OMPL [6] the state of the art software in motion
planning for manipulation. The different tutorials are briefly explained hereafter.
4.1 Motion planning in joint space
This is the basic tutorial of motion planning. It shows how to define the kinematic
configuration to reach in joint space and how to generate a plan and execute it
in MoveIt!. This is a good example on how to obtain collision-free and joint limit
avoidance when moving the robot upper body in joint space.
11
http://moveit.ros.org
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 35
35
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
7
Fig. 4. AMCL-based localization of TIAGo and path planning
4.2 Motion planning in cartesian space
This tutorial allows the user to define the goal kinematic configuration in carte-
sian space, i.e. by defining the goal pose of the given frame of the robot’s upper
body. Then, it is an example on how to run inverse kinematics using MoveIt!
4.3 Motion planning taking Octomaps into account
This tutorial is an extension of the previous one in order to add vision-based col-
lision avoidance with the environment. Indeed, it is an example on how MoveIt!
can handle a 3D representation of the environment based on Octomap’s occu-
pancy grids [7], see 5.
Fig. 5. Motion planning including Octomap occupancy grid
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 36
36
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
8
4.4 Demonstration of a pick & place application
The last tutorial of motion planning includes a full pipeline to have TIAGo
picking an object from a table, raising it and then place it back to its initial
position. The tutorial shows an example on how to use monocular model-based
object pose estimation and how to use MoveIt! to perform the pick and place
operations by avoiding collisions with the table. A snapshot of the demo is shown
in Figure 6.
Fig. 6. Pick and place demo using MoveIt!
5 Perception
5.1 2D perception
A collection of tutorials based on OpenCV [8] are included in order to show
typical tasks in robotics involving computer vision. These tutorials cover the
following areas:
– Color-based tracking
– Keypoint detection and descriptors computation
– Fiducial markers detection and pose estimation
– Person and face detection
– Planar textured object detection based on homobraphy estimation
Figure 7 shows a couple of simulations. On top, the output of the tutorial
showing how to perform person detection. On the bottom, the example of tex-
tured planar object detection and pose estimation.
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 37
37
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
9
Fig. 7. Some examples of 2D perception in simulation
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 38
38
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
10
5.2 Point cloud perception
TIAGo is provided with a RGBD camera placed in its head. Nowadays this kind
of low cost depth sensors are of common use in robotics applications. Further-
more, the existance of the Point Cloud Libray [9] provides a useful set of tools
to process the data from the depth sensors and obtain semantic data out of it.
A set of tutorials showing point cloud processing are presented. The tutorials
aim to show how to detect, segment and estimate the pose of and object on top
of a table for robotic manipulation tasks. Figure 8 show the result of the tutorial
providing cylindrical object detection and pose estimation.
Fig. 8. Table top cylindrical object detection and pose estimation
The tutorials show step by step how to obtain this result starting from the
detection of the plane of the table, following with the segmentation of the point
clusters on top of such a plane, and finally identifying the clusters that fit on
the cylindrical model.
The result of these tutorials can be then combined with the motion planning
tutorials in order to implement grasping tasks with integrated perception.
6 Conclusions
This paper presents the on-line tutorials of ROS for the mobile manipulator
TIAGo. The paper has shown how this tutorials can be used as practial examples
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 39
39
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)
11
when teaching fundamental topics in Robotics and how they can be used as
demonstrators of the different theoretical approaches involved.
The authors are convinced that these simulation-based tutorials can reduce
the learning curve of robotics students as well as being powerful tools for teachers
in order to engage easily their students by providing them with a framework
where practical results are achieved very fast.
Since the publication of the tutorials, on early October 2016, a number of
users, from teachers to PhD/under-graduate students, have adopted them to
teach or learning some aspects of robotics. One important feedback provided by
some of the users was that the installation part of the tutorials was a bit hard
to meet in all existing Operating Systems and/or versions of ROS. In order to
partially overcome this limitation the tutorials were also published in the one-
line site of The Construct12 , so that the tutorials can be followed just by using
a Web browser. The on-line course also offers exercises to practise the different
topics taught in the tutorials.
References
1. Quigley, M., Conley, K., Gerkey, B., Faust, J., Foote, T., Leibs, J., Wheeler, R., Ng,
A.Y.: ROS: an open-source robot operating system. In: ICRA Workshop on Open
Source Software. Volume 3., Kobe (2009) 5
2. Chitta, S., Marder-Eppstein, E., Meeussen, W., Pradeep, V.,
Rodrı́guez Tsouroukdissian, A., Bohren, J., Coleman, D., Magyar, B., Raiola, G.,
Lüdtke, M., Fernández Perdomo, E.: ros control: A generic and simple control
framework for ros. The Journal of Open Source Software (2017)
3. Davidson, J., Hunt, K.: Robots and Screw Theory: Applications of Kinematics and
Statics to Robotics. Oxford University Press (2004)
4. Thrun, S., Fox, D., Burgard, W., Dellaert, F.: Monte carlo localization for mobile
robots. In: In Proceedings of the IEEE International Conference on Robotics and
Automation (ICRA. (1999)
5. Chitta, S., Sucan, I., Cousins, S.: Moveit![ros topics]. IEEE Robotics & Automation
Magazine 19(1) (2012) 18–19
6. Şucan, I.A., Moll, M., Kavraki, L.E.: The Open Motion Planning Li-
brary. IEEE Robotics & Automation Magazine 19(4) (December 2012) 72–82
http://ompl.kavrakilab.org.
7. Wurm, K.M., Hornung, A., Bennewitz, M., Stachniss, C., Burgard, W.: Octomap:
A probabilistic, flexible, and compact 3d map representation for robotic systems.
In: In Proc. of the ICRA 2010 workshop. (2010)
8. Bradski, G.: The OpenCV Library. Dr. Dobb’s Journal of Software Tools (2000)
9. Rusu, R.B., Cousins, S.: 3d is here: Point cloud library (pcl. In: In Robotics and
Automation (ICRA), 2011 IEEE International Conference on, IEEE 1–4
12
http://www.theconstructsim.com
TRROS 2018 – European Robotics Forum 2018 Workshop “Teaching Robotics with ROS”
TRROS 2018
(Edited by S. –Schiffer,
European A. Robotics Forum
Ferrein, M. 2018 Workshop
Bharatheesha, “Teaching
and C. Robotics
Hernández with ROS”
Corbato) 40
40
(Edited by S. Schiffer, A. Ferrein, M. Bharatheesha, and C. Hernández Corbato)