Introduction to ROS

A. Angelou, Prof. Ioannis Pitas

Aristotle University of Thessaloniki
[email protected]
www.aiia.csd.auth.gr
Version 1.0
Date: 8/12/2020
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

2
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

3
What is ROS?

• ROS stands for “Robotic Operating System”

• It’s not an operating system, but a development tool
• Runs through Linux
• Is Open Source
• Supports C++ and Python programming languages
ROS Applications
ROS is used
• For research purposes
• In Research and Development (R & D) Departments in Industry
• By individuals for personal projects

ROS can be used in a wide range of applications such as:

• Autonomous Driving
BlueROV2 [BLRV] IRB 120 Robot [ABB]
• Controlling Robotic Arms
• Drones
• Object Detection
• Object Tracking
• Object Recognition
• Gesture Recognition
• Control of Multi-Drones
ROS Software
• ROS is an open source software and can be installed in any
computer with Linux operating system.
• ROS 2 is supported by Windows 10, MacOS and Linux.
• Each ROS distribution is supported by a specific Linux Distribution.
• ROS has tools, libraries and drivers for both C++ and Python
programming languages that helps you out to develop your
application.
• Has a large community
• It is well supported. Provides tutorials, documentation, libraries etc
• Events (RoboCon, ROS-Industrial Conference, ROS Summer
School)

6
ROS Distributions
The most stable and recent ROS Distributions are:
• ROS Melodic Morenia (Ubuntu 18.04 - Bionic Beaver)
• ROS Noetic Ninjemys (Ubuntu 20.04 - Focal)

http://wiki.ros.org/Distributions 7
ROS Hardware
For ROS application can be used a variety of computer boards:
• Raspbery Pi (Raspberry Pi 4 B)
• PC motherboards (Ashrock X570 Extreme4)
• Embedded motherboards (Nvidia Jetson Nano)

Raspberry Pi 4 ASHROCK PC Motherboard NVIDIA Jetson Nano

[RASP] [ASHR] [NVID]
8
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

9
Robot Architecture
❑ A Robot may have a variety of hardware devices that a developer has to program
and control such as
• Sensors (Altimeter, LIDAR, Gyroscopes, Humidity Sensors etc.)
• Motors (Brushless Motors, AC Motors, Stepper Motors etc.)
• Displays (LCD Display, TFT etc.)
• Communication Devices (GPS/GSM, Bluetooth, Wifi, IR)

❑ But also may develop algorithms necessary for

• Processing Data
(Calculations, Filters, Object Detector, Image Segmentation etc)
• Optimization
(PID Controller, Bee-Hive Algorithm, Wolf-Pack Algorithm, Differential Evolution
Algorithms etc.)

10
Robot Hardware Architecture
Example of Drone
Camera MOTOR 1
Hardware
Block Diagram LIDAR MOTOR 2

MOTOR 3

Control Unit MOTOR 4

Gyroscope 1
MOTOR 5

Gyroscope 2
Altimeter MOTOR 6

Gyroscope 3
GSM/GPS
7

11
Robot Software Architecture
Example of Drone Software Block Diagram

Object Detector
Algorithm DISPLAY
Driver
Object Tracker
Camera Driver Algorithm

GIMBAL Driver

GIMBAL Control

12
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

13
ROS Architecture
❑ We can describe ROS in two levels:
• File System level
• Computational level

A] File System Level

How ROS Directories, Folders and files are organized in the system.

B] Computational Level
How ROS programs communicate with each other.

14
A) ROS Computational Level
The basic ROS concepts are:
• Nodes

• Topics

• Messages and Bags

• Services

• Packages

• ROS Master

• ROS Core

• ROS Tools

15
Nodes
❑ A Node is a piece of code that performs a specific function to the robot.
❑ They are simple programming files with the extension
• .py (Python)
• .cpp (C++)

A Robot usually consist of many nodes, such as

• Camera control
• LIDAR
• GIMBAL Control
• Display
• Object Detector
• Object Tracker
• Image Filter
• Etc.

16
Messages and Bags
MESSAGES
ROS uses messages so as nodes can sent or receive data. Messages contains Fields
and Constants.

• Fields
Fields contain the data. The supported data types are integer, float, boolean,
arrays, structures etc.

• Constants
Constants are values that can used to interpret the fields.

BAGS
A bag contains many messages.

http://wiki.ros.org/Messages 17
Topics
❑ Topics are names that are used from the Nodes to communicate with each other
and exchange messages.
❑ A topic is associated with the message type and thus only nodes with the same
type can send or receive messages. For example, a Node that sends floating point
data to a topic, the data cannot be received by a boolean type Node.

Alarm
TOPIC Node
Type : Boolean
Chemical Chem_Data
Sensor Node Type: Floating Point
14.36905 Plot
Node
Type: Floating Point

18
Services
A Publish/Subscriber model is a very convenient one-way
communication between nodes. In distribution systems, there’s a need
for data exchange between nodes in a two-way direction. For this
reason the Service model is used.

A Service is a request/reply model for node communication using two

types of messages:
• A Request message
• A Reply message

Like messages, a service type depends on the data type.

19
 Services

Service Server Service Service

Node Name Client
Type: Floating Node
Point

20
Packages
Packages
Packages are a set of nodes, manifests, libraries, files (pictures,
videos, data), code pieces that are gathered together in a folder
so it can be reused easily.

Packages have a specific file structure when they are created. That
doesn’t mean that the developer cannot create his own folders and files
inside the package folder.

21
ROS Master
❑ The ROS Master is the coordinator of the communication between
nodes.

❑ All Nodes, Topics Services are registered to ROS Master.

❑ When a Node wants to sent a message to a Topic or exchange
messages with the another Node, ROS Master provides a way to the
Nodes to locate each other.
❑ After the Nodes identify each other, they are communicating

22
 ROS Master
When an Node wants to publish a message to a Topic, the Publisher Node notify ROS Master to send
data to the Topic.

ROS MASTER

TOPIC
Plot
Chemical Chem_Data Node
Sensor Node Type: Floating Point Type: Floating Point

23
 ROS Master
After the notification, the Publisher Node establishes connection the Topic. At this point, the publisher
doesn’t sent any message to the Topic unless a Subscriber Node notify ROS Master.

ROS MASTER

TOPIC
Plot
Chemical Chem_Data Node
Sensor Node Type: Floating Point Type: Floating Point

24
 ROS Master
When an Subsciber Node wants to subscibe a message from a Topic, the Subsciber Node notify ROS
Master to connect to the Topic.

ROS MASTER

TOPIC
Plot
Chemical Chem_Data Node
Sensor Node Type: Floating Point Type: Floating Point

25
 ROS Master
• After the notification, the Subsciber Node connects to the Topic.
• At this point the Publisher Node publishes the data to the Topic and the Subsciber, subscibes to the
Topic.
• The data is transmitted from the Publisher Node to the Subsciber Node through the Topic.

ROS MASTER

TOPIC
Plot
Chemical Chem_Data Node
Sensor Node Type: Floating Point Type: Floating Point

26
ROS Core
❑ ROS core is a collection of routines,
nodes, libraries that are essential for
ROS system ROS Master

❑ It runs at the background. ROS Parameter

Server

❑ ROS Core starts the ROS Master to rosout

enable the registration of all Nodes,
Topics and Services.

27
ROS Core

28
ROS Tools
ROS provides a variety of tools to build, debug and
simulate . The Most common tools are:

• Catkin

• rqt_graph

• Opencv Library

• Gazebo

29
Catkin
What is Catkin?

❑ Catkin is a tool that is included with ROS and it is used to

build packages.
❑ The name Catkin was given by the Willow Garage
Company that created ROS.
❑ It was created for easy package installation and
distribution.
❑ It consist of macro instructions and scripts to build
packages Image of male Catkin
[CTKN]

30
Rqt graph
Rqt_graph is GUI tool that shows the function of all nodes and topics of a
ROS project.

A typical rqt_graph showing the nodes and topics at a graph level

[RQTG]

31
OpenCV Library
OpenCV is an open source library for computer vision, machine
learning and real-time applications. The library includes
functions for:

• Object Detection
• Deep Neural Networks
• Machine Learning
• Image Processing
• Video Analysis
• 3D Reconstruction with Camera
• Image or Video Input and Output

32
Gazebo
Gazebo is a simulator for testing and training robots using
realistic scenarios in virtual environments

A simulation of a scenario with various robots in Gazebo

[GZBO]
33
B) File System Level

The image is from the paper [RRSham]

34
Catkin Workspace Folder Location

35
Catkin Workspace

36
Catkin Build folder

37
Src Folder with Packages

38
Package Folder

39
Node Folder

40
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

41
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

42
Publisher and Subscriber Nodes
Assume that we receive a n image from a camera and we want to show
this image. We must create a Publisher Node, a Topic and a subscriber
Node.

TOPIC
Image
Image Subscriber
Publisher Node Image
Node

43
Publisher Node

44
Subscriber Node

45
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

46
Object Detector and Tracker Node

47
Example Structure

48
Example Structure

49
Object Detector and Tracker

50
Presentation Outline
• What is ROS
• Robot Architecture
• ROS Architecture
o Computational Level
o File System Level
• Examples
o Publisher and Subscriber Nodes
o Object Detector and Tracker Node
o Object Detector Node

51
Object Detector Node

2
#! /usr/bin/env python
3

4
import rospy
5
from std_msgs.msg import Header
6
from sensor_msgs.msg import CompressedImage
7
from detector.msg import BoundingBox
8
from std_srvs.srv import Empty , EmptyResponse
from keras.applications.imagenet_utils import preprocess_input
9

10
from keras.preprocessing import image
11
import cv2
12
import tensorflow as tf
from deep_learning.ssd_detector.ssd import SSD300 as SSD
Object Detector Node
1 class DetectionPoseNode :
2 def __init__(self) :
3 models_path = join(os.path.dirname(os.path.realpath( __file__) ) , ’ models ’ )
4 self.pub_bbox = rospy.Publisher(’face_detector/bbox ’ , BoundingBox ,
5 queue_size = 10)
6 self.class_names = [ ” background ” , ’face’]
self.num_classes =len(self.class_names)
8 self.input_shape = (300 , 300 , 3)
7 self.conf_thresh = 0.6
9
10
11 self.model = SSD (self.input_shape , num_classes = self. num_classes )

12 self.model.load_weights (join(models_path , ’ fddb_model. hdf5 ’ ) )

13 self.pose_model = Pose_Estimator(join(models_path ,
‘pose_estimation_model.h5’))
14 self.bbox_util = BBoxUtility(self.num_classes )
15 self.graph = tf.get_default_graph( )
Object Detector Node

1 def listener(self) :
2 rospy.init_node(’detector’ , anonymous=True )
3 rospy.loginfo(’Detector node started!’)
4 rospy.Subscriber(”/usb_cam/image_raw/compressed ” ,
CompressedImage, self.detection_callback , queue_size =1)

5 rospy.spin( )
Object Detector Node

1 def detection_callback(self, data) : ,

2 # Compressed image
3 np_arr = np.fromstring(data.data , np.uint8)
4 cv_image = cv2.imdecode (np_arr , cv2.IMREAD_COLOR)
5 # For uncompressed images we need to use the cv2 bridge
6 # cv_image = self.bridge.imgmsg_to_cv2( data , ” bgr8 ” )
7 self.process_image(cv_image)
Object Detector Node

1 if __name__ == ’ __main__ ’ :
2 node = DetectionPoseNode( )
3 node.listener( )
Object Detector Node

1 def process_image(self , orig_image) :

2 resized = cv2.resize(orig_image, (self.input_shape[0] , self.input_shape[1]))

3 rgb = cv2.cvtColor(resized, cv2 .COLOR_BGR2RGB)

4
5
6 # Use model to predict
7 with self.graph.as_default( ) :
8 x = preprocess_input(np.array([image.img_to_array(rgb) ]))

9 y = self.model.predict(x)
10 results = self.bbox_util.detection_out(y)
11
12 timestamp = rospy.get_rostime( )
Object Detector Node

1 if len(results) > 0 and len(results[0]) > 0 :

2 det_label , det_conf, det_xmin = results[0][ : , 0 ] ,
results[0][: , 1 ] , results[0][: , 2 ]
3 det_ymin , det_xmax , det_ymax = results[0][: , 3 ] ,
results[0][: , 4 ] , results[0][: , 5 ]
4
5 top_indices = [ i for i , conf in enumerate(det_conf) if
conf >= self.conf_thresh]
6
7 top_conf = det_conf[top_indices]
8 top_label_indices = det_label[top_indices]. tolist()
9 top_xmin , top_ymin , top_xmax , top_ymax = det_xmin [
top_indices] , det_ymin [top_indices], det_xmax [
top_indices], det_ymax [top_indices]
Object Detector Node

1 for i in range ( top_conf.shape[0] ) :

2 xmin , ymin = top_xmin [ i ] , top_ymin [ i ]
3 xmax , ymax = top_xmax [ i ] , top_ymax [ i ]
4 bbox = BoundingBox ( header=Header ( stamp=timestamp ) ,
x_min=xmin , x_max=xmax , y_min=ymin , y_max=ymax)
5 self.pub_bbox.publish( bbox )
References
[BLVR] https://bluerobotics.com/store/rov/bluerov2
[ABBR] https://new.abb.com/products/robotics/industrial-robots/irb-120
[RASP] https://www.raspberrypi.org/products
[ASHR] https://www.asrock.com/mb/AMD/X570%20Extreme4/index.asp
[NVID] https://developer.nvidia.com/embedded/jetson-nano-developer-kit
[CTKN] https://upload.wikimedia.org/wikipedia/commons/8/83/Ostrya_carpinifolia_in_Italy_male_
catkins.jpg
[RRSham] Redmond R. Shamshiri, Ibrahim A. Hameed, et.al, “Robotic Harvesting of Fruiting Vegetables: A
Simulation Approach in V-Rep, ROS and MATLAB”, Automation in Agriculture - Securing Food Supplies for
Future Generations, InTechOpen, Feb, 2018
[RQTG] https://roboticsbackend.com/rqt-graph-visualize-and-debug-your-ros-graph
[GZBO] https://en.wikipedia.org/wiki/File:Gazebo_screenshot_v5.0.png
[BLVR] https://bluerobotics.com/store/rov/bluerov2
[ABBR] https://new.abb.com/products/robotics/industrial-robots/irb-120

60
References
References for Further Reading

Book 1 : “Robot Operating System for Absolute Beginners, Robotics Programmings Made Easy”, Lentin
Joseph, Apress, 2018
Book 2 : “Robot Operating System – The Complete Reference”, Anis Koubaa, Springer, Vol 1, 2016
Book 3 : “Robot Operating System – The Complete Reference”, Anis Koubaa, Springer, Vol 2, 2017
Book 4 : “ROS By Example, A Do-it Yourself Guide to the Robot Operating System”, R. Patrick Goebel,
Vol 1, 2012

ROS Wiki Link : http://wiki.ros.org

OpenCV Link : https://opencv.org
Gazebo Link : http://gazebosim.org
Robots with ROS: https://robots.ros.org

61
 Q&A

Thank you very much for your attention!

Contact: Prof. I. Pitas

[email protected]

62