Author Archive Ryosuke Tajima

ByRyosuke Tajima

Translation of ROS-I training material

What is ROS-Industrial?

When examining the combination of robot arm such as industrial use and ROS, “ROS-Industrial (ROS-I)” is found. Does it support my robot or my application? What is ROS-I, is it different from ROS? We have received many such inquiries at TORK.

ROS-I is the name of an international consortium . In this consortium, the participating member companies share technical problems in using ROS for industrial use, and companies in the member provide packages and solutions (based on the contract). ROS-I standards and packages are not independent of ROS. After all, it is a story in the ROS ecosystem.

Packages developed and managed by ROS-I will be shared only within member companies for a while, but in the end it will be open source under the contract. The ROS-I repository is the result of these.

• ROS-Industrial repository

ROS-I regularly holds training courses to use this result in North America, EU, and Asia in Singapore. It is about 3 days course. This training course can be taken even if you are not a member of ROS-I.

• ROS-Industrial events

Furthermore, the training materials for the training are also made public, and anyone interested if referring to this material can study the main packages of ROS-I.

• ROS Industrial Training Exercises

In addition to the basic commentary and setup of ROS, it is also possible to write a node in C++, MoveIt!, a recently developed orthogonal coordinate planning package “Descartes” in ROS-I, and point cloud processing, image processing, and so on. The exercise is mainly the style of experimenting with your PC or VM.

To be honest, I think that the difficulty is rather high and not for beginners. But If you’ve already learned ROS and going to use the industrial robot from now on, it seems to be a very helpful reference.

TORK translated the training material into Japanese!

“But it is in English…” There is good news for you! We translated the material of this training course into Japanese, and was completed at this time!

ROS Industrial training material in Japanese

It would be fun for a challenge for winter vacation.

As a precaution, when there is a problem, mistake, improvement point in the content of the Japanese version , please contact TORK rather than the original authers.

Have a wonderful year!

ByRyosuke Tajima

TORK is participating in B-Boost Conference

TORK is invited to the open source software conference “B-Boost” held in Bordeaux, France on 6 and 7 November . We will talk about robots and ROS.

We will report the situation. Je suis parti.

ByRyosuke Tajima

TORK received “RT Middleware Promotion Contribution Award”

We received “RT Middleware Promotion Contribution Award” from the Japan Robot Industry Association. It is a cooperative award with Kawada Robotics. NEXTAGE Open’s software support was appreciated. We will continue to contribute to open source software in the robot field.

ByRyosuke Tajima

We partecipated in ROS-I Asia Pacific Workshop!

We participated in the ROS Industrial Asia Pacific Workshop held in Singapore on June 27 and 28. (Sorry, this blog have been a bit out of date waiting for the update of the official page.)

• ROS-Industrial AP Workshop 2018: https://rosindustrial.org/events/2018/6/27/ric-ap-workshop

It was a new and clean facility

It is our first participation from TORK. Rather than the technical contents, there were many introductions of projects in each country and efforts by companies, and it was a different impression from ROSCon specialized in ROS package and technology. Instead I was able to know the activities of major overseas companies and venture companies. Also, it was very pleasant to have many opportunities to chat with those people and get acquainted.

TORK also introduced the contents of the activity, the jog_control package and also exhibited the ROS native teaching pendant under development, and many people were interested.

In ARTC there is a very spacious and cool building with the name “Future Factroy”, and many industrial robots were lined up and narrowed down. Demonstration of ARTC ‘s robot team were shown. ROS and the latest industrial robot were combined and research was conducted targeting themes such as Scan-N-Plan and object recognition. Unfortunately it was not allowed to take photos for participants in the workshop, but I think the state will be released following an official blog etc.

In the whole workshop , “everyone has no doubt on the industrial application of ROS” (because it is an event of ROS-I, which is natural). Participants of ROS-I are actually trying to concretely link ROS to application and to strengthen their product capabilities. Companies in China and Taiwan also actively release ROS and ROS 2 products. Although it is certain that there are some challenges in industrial application of ROS,  there are efforts such as package creation support by ROSIN, mechanism of code quality management, development of Scan-N-Plan fundamental technology. Most of all, I felt it was wonderful that they loop around making hands move and publish results, which we want to do as TORK.

And the organizer’s operation was wonderful! I would like to express my thanks to the very fun and comfortable workshop with kind attentions.

ByRyosuke Tajima

We are heading to ROS-I Asia Pacific Workshop!

TORK will participate in the ROS Industrial Asia Pacific Workshop which will be held in Singapore on 27th and 28th June.

• ROS-Industrial AP Workshop 2018

We expect many lectures from ROS-I developers SwRI, Faunhofer IPA, corporate members such as OpenRobotics, ADLINK and Boeing. What kind of story can we hear?

TORK  would like to appeal a wide range of activities such as currently developing packages. I will also report on the contents on this blog.

ByRyosuke Tajima

Low-cost LIDARs(2) Specification

Product Specifications

In the previous post, we introduced low-cost LIDARs.
Let’s compare the specifications of the sensors.

• RPLIDAR A2 (Slamtec )
• Sweep (Scanse)

It is an important precaution!

• Be sure to check the specifications by yourself when you actually use it!
• TORK can not take responsibility even if your project does not work out due to errors in this article!

Principle

For industrial LIDAR, distance measurement by ToF (Time of Flight) method is often used. This is a method of calculating the distance by measuring the time until the laser light is reflected by the object and returned to the sensor. Sweep uses the ToF method to measure distance. On the other hand, RPLIDAR A2 uses the principle of triangulation instead of ToF. This is a method of measuring the distance to the reflection point by measuring the displacement of the light receiving point with a sensor. The light emitting point, the reflection point and the light receiving point constitute a triangle, I believe you may know already.

Triangulation generally seems easier in general, but as the distance to the reflection point increases, even if reflected light can be measured, the resolution of distance decreases. On the other hand, the ToF can increase the maximum measuring distance because the resolution does not decrease even if the distance to the reflection point increases (as far as the reflected light can be measured).

There is another big difference. For RPLIDAR A2, sampling is done in synchronization with the angle. In other words, the angle of the laser to be measured is the same for each rotation (even if there are delicate variation actually). On the other hand, although Sweep is synchronized with the scan start at the angle 0 position, each distance measurement is not synchronized with the angle. From the specification of Sweep, the data obtained from the sensor is a set of angle and distance. The laser that Sweep emits has a pattern rather than a single pulse, so the sample rate is not constant (default is 500 to 600 Hz), so it seems to be like this.

High-end two-dimensional LIDAR such as URG and SICK is a method that can acquire distance synchronized with angle, so it seems that Sweep’s method is a bit difficult to use.

Angle resolution, measurement cycle, sample rate

As an important performance in LIDAR, there is angle resolution and measurement cycle. The angular resolution is the fineness of the angle of the measurement point. When the resolution is low, the measurement points of the object and the environment are too small, and the shape of the measurement is not well understood. The measurement period is the time taken for one scan (one rotation), and when the measurement cycle is large, accurate measurement becomes difficult when measuring a moving object or when the robot itself moves.

Both of the sensors are measuring the surrounding by rotating the part where the light source and the measurement part are integrated. RPLIDAR A2 has variable rotation speed from 5 Hz to 15 Hz, but the number of measurement points is the same (400 points) regardless of the rotation speed.

Sweep has a variable rotation speed from 1 Hz to 10 Hz, but since the sample rate is 1000 samples / sec at maximum (precisely 1000 to 1025), the faster the rotation speed is, the measurement points per revolution is getting lesser.

• RPLIDAR A2: 400 samples / rotation
• Sweep: 1000 samples / sec

If both are rotating at 10 Hz,

• RPLIDAR a 2: 400 samples / rotation
• Sweep: 100 samples / rotation

Therefore, RPLIDAR A2 has an angular resolution four times higher. If you want to obtain the same resolution with Sweep, you need to rotate the cycle at 1/4 (2.5 Hz).

Maximum measurement distance and distance resolution

Maximum measurement distance is also important performance. It seems better to measure distantly, however, from the experience 10m is enough for office navigation, and 100 m will not be enough for  applications like automatic driving. It depends the application.

RPLIDAR A2, confusingly, the measurement distance is different depending on the model. Attention is required for model number when purchasing. This time I used RPLIDAR A2M8 with a measuring distance of up to 8 m.

• RPLIDAR A2M4 Maximum measurement distance 6 m
• RPLIDAR A2M8 Maximum measurement distance 8 m
• RPLIDAR A2M6 Maximum measurement distance 16 m

On the other hand, Sweep has a maximum measurement distance of 40 m.

By the way, the minimum measurement distance, which is the limit of the near side, is about the same.

• RPLIDAR A 2 M 8 minimum measurement distance 0.15 m
• Sweep Minimum measurement distance 0.15 m

How about distance resolution? The distance resolution of RPLIDAR A2M8,

• 0.5 mm or less when 1.5 m or less
• At a distance more than 1.5m, 1% of the distance or less

The nearer the resolution is higher, so the shape of the object can be taken properly if it is close. Instead, it can be predicted that both distance and resolution will fall as the distance increases (data did not exist for errors in specifications).

On the other hand, for Sweep, 

• Sweep distance resolution 1cm or less

regardless of the measurement distance. The Sweep specification also includes a graph of the distance measurement error, but you can see that the error does not increase even if the measurement distance becomes longer. It is an advantage of ToF method.

Sweep measurement error (excerpted from Sweep manual)

Others

Both are fairly small, but since the top parts rotate, installation is necessary to avoid touching them.

• RPLIDAR A2: diameter 70 mm
• Sweep: Diameter 65 mm

For Sweep, the cable connector is oriented in two directions, one side and the lower side, making it easy to wire. RPLIDAR seems to be difficult to shorten the cable length because the cable is straight out.

Connector on the back side of Sweep

When both are connected by USB, they operate with USB bus power. Also, you can communicate with a UART such as a microcomputer, but in that case you need a separate power supply.

Summary

Although it is two sensors whose outline looks the same, you can see that the design philosophy is actually slightly different. It can be said that RPLIDAR A2 is short distance, for indoor use, and Sweep is specification for long distance and outdoor use. I am afraid that angle resolution of Sweep might be low for some usages.

Next we see the ROS support for both.

ByRyosuke Tajima

New aibo has arrived!

Sony’s pet robot “aibo” has arrived in TORK!

We named it “TORK-chan (♀)”.

They announced “ROS are mounted in aibo”, and it seems real. Cool!

aibo まじでros-kineticだった．https://t.co/V1FlT3He9l

衝撃

— OTL (@OTL) January 11, 2018

You can see in the following pages the list and licensing of open source software that are installed.

• http://oss.sony.net/License/ERS-1000/

In addition, we can now download the open source codes from here.

• http://oss.sony.net/Products/Linux/EntertainmentRobot/category01.html

While we examined a little, we can not find the way to access the aibo system from the user.  We can expect some SDKs will be published in near future. Until the time being, let’s enjoy this cute pet robot!

ByRyosuke Tajima

Interactive ROS Indigo environment with Docker (1)

Indigo? Kinetic?

ROS will continue to be newly released. The latest version is Lunar Loggerhead (Lunar) and the previous, Kinetic Kame (Kinetic) and  Jade Turtle (Jade). Traditionally, they named after turtles.

• ROS – Distributions

However, I think still many people are actually using the previous version Indigo Igloo (Indigo). There are many Indigo packages that have not yet been released in Kinetic. For the same reason, even TORK is currently using the Indigo in our workshop materials and contract development .

Indigo is based on Ubuntu14.04 while Kinetic and Lunar is based on Ubuntu16.04 or later. We feel a bit outdated to install Ubuntu14.04 into our PC only in order to use the Indigo.

It could be a solution to use virtual machine (VM), such as VMWare or VirtualBox.  However, it can be a problem when you try to use the hardware. For example, there are often device driver problem such as graphics performance, USB camera compatibility, not compatible with USB3.0 and so forth.

Docker can be the remedy for this issue. You can work easily in the interactive Indigo development environment using devices and GUI while you installed the Ubuntu16.04 on your PC. 

Installation of Docker

First install Docker. Let’s install the latest version according to the instructions in the pages of the Docker. Use of that Community Edition (CE). 

• Get Docker CE for Ubuntu

Preparation of Indigo of Docker image

OSRF prepares the Docker container of Indigo on DockerHub.

• Docker Hub – osrf

Get the Indigo of the container with the following command.

$ docker pull osrf/trusty-ros-indigo:desktop

By the way, TORK also publishes the Indigo Docker container. It contains additional packages for the TORK workshops, Nextage Open and other useful tools. You may also find it here.

• Docker Hub – tork/indigo

For example, if you want to check the version by the command ‘lsb_release’, 

$ docker run osrf/trusty-ros-indigo:desktop lsb_release -a

No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 14.04
LTSRelease: 14.04
Codename: trusty

User permissions of execution will be root by default. To change the user rights and the access rights to the devices, you need to set the options at the time of the separate docker run. For more information around here, Please refer to Docker Run Reference.

Interactive start-up that share user and home directory

Here we want to run the shell with their own user rights, and to work interactively in the home directory. We created a script to do some of the settings that are required for that.

• github: docker_shell

$ ./docker_shell osrf/trusty-ros-indigo:desktop

After that, you are on the shell as if it’s OS has become Trusty + ROS Indigo.

$. /opt/ros/indigo/setup.bash

You can use the shell like normal Indigo environment. You must be careful the home directory is shared. If you are using the Kinetic in the base system, you can easily get confused whether your workspace is for Kinetic or Indigo.

If you are using the NVidia driver

If the PC is not using a special X graphics driver, you can also use GUI, such as rviz. But if the PC is using the NVidia driver, it does not work apps, including OpenGL, such as rviz. This solution will introduce the next time.

ByRyosuke Tajima

ROS + Snappy Ubuntu Core (3) ： Looking into the Snappy package

How does the Snappy package, we created in the last blog post works?

$ which talker-listener.listener 
/snap/bin/talker-listener.listener

$ ls -l /snap/bin/talker-listener.listener 
lrwxrwxrwx 1 root root 13 Jul 21 16:04 /snap/bin/talker-listener.listener -> /usr/bin/snap

This seems to be equivalent to doing the following.

$ snap run talker-listener.listener

The snap file is simply expanded to /snap/. The directory ‘current’ should the latest version, but this is a symbolic link to ‘x1’. It will be easy to roll back.

$ ls -l /snap/talker-listener/
total 0
lrwxrwxrwx  1 root root   2 Jul 21 16:04 current -> x1
drwxrwxr-x 10 root root 222 Jul 21 16:00 x1

$ ls /snap/talker-listener/current
bin                       command-roscore.wrapper  etc  meta  snap  var
command-listener.wrapper  command-talker.wrapper   lib  opt   usr

I see the command-listener.wrapper. Looking inside,

#!/bin/sh
export PATH="$SNAP/usr/sbin:$SNAP/usr/bin:$SNAP/sbin:$SNAP/bin:$PATH"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$SNAP/lib:$SNAP/usr/lib:$SNAP/lib/x86_64-linux-gnu:$SNAP/usr/lib/x86_64-linux-gnu"
export ROS_MASTER_URI=http://localhost:11311
export ROS_HOME=${SNAP_USER_DATA:-/tmp}/ros
export LC_ALL=C.UTF-8
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$SNAP/lib:$SNAP/usr/lib:$SNAP/lib/x86_64-linux-gnu:$SNAP/usr/lib/x86_64-linux-gnu
export PYTHONPATH=$SNAP/usr/lib/python2.7/dist-packages:$PYTHONPATH
export PATH=$PATH:$SNAP/usr/bin


# Shell quote arbitrary string by replacing every occurrence of '
# with ''', then put ' at the beginning and end of the string.
# Prepare yourself, fun regex ahead.
quote()
{
    for i; do
        printf %s\n "$i" | sed "s/'/'\\''/g;1s/^/'/;$s/$/' \\/"
    done
    echo " "
}

BACKUP_ARGS=$(quote "$@")
set --

if [ -f $SNAP/opt/ros/kinetic/setup.sh ]; then
    _CATKIN_SETUP_DIR=$SNAP/opt/ros/kinetic . $SNAP/opt/ros/kinetic/setup.sh
fi

eval "set -- $BACKUP_ARGS"

export LD_LIBRARY_PATH="$SNAP/opt/ros/kinetic/lib:$SNAP/usr/lib:$SNAP/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH"
export LD_LIBRARY_PATH=$SNAP_LIBRARY_PATH:$LD_LIBRARY_PATH
exec "rosrun" roscpp_tutorials listener "$@"

It seems that the command is executed after ROS environment setting etc are done with this wrapper file. In other words, you do not need “source setup.bash” which you always do with ROS. It seems to be easier.

What else is included in the snap package?

$ tree -L 3 -d /snap/talker-listener/current/
/snap/talker-listener/current/
├── bin
├── etc
│   ├── ca-certificates
│   │   └── update.d
│   ├── dpkg
│   │   └── dpkg.cfg.d
│   ├── emacs
│   │   └── site-start.d
│   ├── gss
│   │   └── mech.d
│   ├── ldap
│   ├── openmpi
│   ├── perl
│   │   ├── CPAN
│   │   └── Net
│   ├── python2.7
│   ├── python3.5
│   ├── sgml
│   ├── ssl
│   │   ├── certs
│   │   └── private
│   ├── systemd
│   │   └── system
│   └── xml
├── lib
│   └── x86_64-linux-gnu
├── meta
│   └── gui
├── opt
│   └── ros
│       └── kinetic
├── snap
├── usr
│   ├── bin
│   ├── include
│   │   ├── apr-1.0
│   │   ├── arpa
│   │   ├── asm-generic
│   │   ├── boost
│   │   ├── c++
│   │   ├── console_bridge
│   │   ├── drm
│   │   ├── gtest
│   │   ├── hwloc
│   │   ├── infiniband
│   │   ├── libltdl
│   │   ├── linux
│   │   ├── log4cxx
│   │   ├── misc
│   │   ├── mtd
│   │   ├── net
│   │   ├── netash
│   │   ├── netatalk
│   │   ├── netax25
│   │   ├── neteconet
│   │   ├── netinet
│   │   ├── netipx
│   │   ├── netiucv
│   │   ├── netpacket
│   │   ├── netrom
│   │   ├── netrose
│   │   ├── nfs
│   │   ├── numpy -> ../lib/python2.7/dist-packages/numpy/core/include/numpy
│   │   ├── openmpi -> ../lib/openmpi/include
│   │   ├── protocols
│   │   ├── python2.7
│   │   ├── rdma
│   │   ├── rpc
│   │   ├── rpcsvc
│   │   ├── scsi
│   │   ├── sound
│   │   ├── uapi
│   │   ├── uuid
│   │   ├── video
│   │   ├── x86_64-linux-gnu
│   │   └── xen
│   ├── lib
│   │   ├── compat-ld
│   │   ├── dpkg
│   │   ├── emacsen-common
│   │   ├── gcc
│   │   ├── gold-ld
│   │   ├── lapack
│   │   ├── ldscripts
│   │   ├── libblas
│   │   ├── mime
│   │   ├── openmpi
│   │   ├── pkgconfig
│   │   ├── python2.7
│   │   ├── python3
│   │   ├── python3.5
│   │   ├── sasl2
│   │   ├── sbcl
│   │   ├── ssl
│   │   ├── valgrind
│   │   └── x86_64-linux-gnu
│   ├── sbin
│   ├── share
│   │   ├── aclocal
│   │   ├── applications
│   │   ├── apps
│   │   ├── apr-1.0
│   │   ├── bash-completion
│   │   ├── binfmts
│   │   ├── boostbook
│   │   ├── boost-build
│   │   ├── bug
│   │   ├── ca-certificates
│   │   ├── cmake-3.5
│   │   ├── debhelper
│   │   ├── dh-python
│   │   ├── distro-info
│   │   ├── doc
│   │   ├── doc-base
│   │   ├── docutils
│   │   ├── dpkg
│   │   ├── emacs
│   │   ├── glib-2.0
│   │   ├── icu
│   │   ├── libtool
│   │   ├── lintian
│   │   ├── man
│   │   ├── mime
│   │   ├── mpi-default-dev
│   │   ├── numpy
│   │   ├── openmpi
│   │   ├── perl
│   │   ├── perl5
│   │   ├── pixmaps
│   │   ├── pkgconfig
│   │   ├── pyshared
│   │   ├── python
│   │   ├── python3
│   │   ├── sgml
│   │   ├── sgml-base
│   │   ├── xml
│   │   └── xml-core
│   └── src
│       └── gtest
└── var
    └── lib
        ├── sgml-base
        ├── systemd
        └── xml-core

144 directories

Wow!

The talker-listener ‘s snap package contains the entire directory structure of the required Linux system as well as ROS. It feels a little overkill… But this is nowadays popular as Docker and other do, so-called container virtualization.

By the way, the generated snap file,

$ ls -sh talker-listener_0.1_amd64.snap 
154M talker-listener_0.1_amd64.snap

There are also 154 Mbyte! Is it acceptable when the package increases?
There are probably ways to solve this because the package contains a lot of unnecessary files.

In the next post, we’ll show how to install Ubuntu Core into the Raspberry Pi 3 machine.