MIKRIK ROS2 Robot on Intel Robotics SDK

Read Part 2 of the tutorial here

Glad to notify, that my robot became available on Tindie. If you are not willing to follow all setup steps and just want a ready-to-run robot for your ROS2 adventures, please support me here.

Slack Support Channel

Met an issue? Please join my Slack channel or leave comments below the post.

Story

Recently I was playing with robots using ROS1 and ROS2. I tried to build an F1tenth robot project https://f1tenth.org using Intel NUC, then switched to a robot vacuum cleaners, played with omnidirectional robots.

I find out that except JetBot by Nvidia, there are not so many x86 powered robots that you can build from A to Z just following an instruction manual. Either you have to buy expensive platform like iRobot Create, Husarion, TurtleBot or just has a risk to be stucked in a ROS1/ROS2 building tutorials mess all over the web.

I decided to find a way to make life of a typical x86 ROS-maker easier. For Intel-based processors I found on the web a Robotics SDK. Literally, it is a set of debian packages that you install on your Ubuntu 22.04 system, and can access Intel-proprietary 3D-vision algorithms like CollabSLAM, ADBSCAN, KudanSLAM. I decided to give it a try, and that is how my new robot project got born.

MIKRIK - is an open-source two-wheel drive robot for robotic enthusiasts who is studying robot vision using 3D-cameras.

Abbreviation of the MIKRIK:

• MIni
• Kreative
• Robotics
• Intellective
• Kit

MIKRIK supports any control board you have at home. It has to run ROS1 on Raspberry to read encoder data, control motors, and creates a ROS1 interface. Then it can be accessed by ROS1/ROS2 navigation node running on a separate computer using ROS1-ROS1 bridge.

Why do I need to use ROS1-ROS2 bridge?

Raspberry is not powerful enough to handle vision part, and Robotics SDK is not supported on ARM architecture. For MIKRIK project it is used only for the low-cost tasks. The high-load computer vision tasks, SLAM, and navigation will be done by a separate board. In my case I will use x86 board LattePanda.

Why don't you use a GPIO of the LattePanda, Radxa, UP7000 or any other board?

The project evolved from Raspberry, and I decided not to spend time writing GPIO support for any board that has GPIO pinout. If you are willing to do so, let me know. Now I'm thinking to switch from Raspberry to OrangePI to cut costs.

Project has two parts:

- Robot part. It runs ROS1 using Raspberry PI 4B. ROS1 has nodes to communicate with the robot motors and to read encoder data. There we will run roscore ROS1 master.

Then it creates a robot controller, after that your robot can be controlled by /cmd_vel topic.

By finishing that part you will have a ROS1 robot to be controlled using a PS4 gamepad.

To go further, you will have to connect it to the client part of the project, and control robot by sending control commands to the /cmd_vel topic.

You might skip control part of the project and enjoy just a solo ROS1 robot based on Raspberry. But, unfortunately, you have to add ROS1 nodes to read Lidar data, and create map using SLAM algorithms like SLAM-ToolBox.

As my tutorial is more focused on using Intel Robotics SDK for computer vision, I will not cover ROS1 and SLAM using Lidars, it is in a To-Do list.

- Control computer part. It is the main brain of the robot. It will generate control commands, and move robot according to the main navigation node program. It can be any powerful enough computer running ROS2 ROS1 SLAM and navigation nodes. Navigaton node will subscribe to the /cmd_vel topic and will publish value to your robot using ROS1-ROS2 bridge.

Your robot can run ROS2 on Intel NUC, powered by x86 Intel CPU board like LattePanda, Radxa X4, or UP7000 board. Because they are all x86, the setup will be identical for all of them. The only thing, maybe some of the boards will require external Wi-Fi module.

Then connect computer to a ROS1 low-level robot part using a ROS1-ROS2 bridge over the Ethernet cable.

In this tutorial I will cover:

• How to build a differential drive two-wheel robot
• How to connect robot to the Intel Robotics SDK and run Visual SLAM using Realsense D435(i) camera

Same way you can connect Lidar, but I strongly recommend you to use 3D-cameras or even better an RGB cameras. RGB camera you can teach using AI to detect depth of the scene, but AI part is not covered in the tutorial, if you're a smart guy, you're welcome to contribute.

Of course for a new makers, I will add a 2D-Lidar support for this project, because 2D-Lidars are cheaper than 3D-cameras, and are good for learning. Wait for the updates.

There might be a slight discrepancy between photos and actual CAD-parts that you have. Project is in the development stage, so CAD-files are getting updated often.

Part 1 Robot Building Steps

If you already have a ROS1 or ROS2 robot, and want just to explore Intel Robotics SDK, you can go straight to Part 2. In Part 2 I will explain you how to set your robot to communicate with Robotics SDK software, and run multiple robotics tutorials from the Developer Guide.

Prerequisites

• Please laser-cut parts from the MIKRIK-CAD Github repo, and 3D-print some of them. Follow the docs on the Github repo for more details.

If you see that some parts are not same as on the photos below, please be aware that I'm updating project constantly, and design can have changes. It doesn't impact robot usability. All navigation ROS2 constants in the code are related to the robot assembly that you can find on Github MIKRIK-CAD Github repo repo.

Photos below are given only for the reference.

Step 1.1 Motors installation

Please take 1 motor mount plate (left), two M2.5x20mm screws with M2.5nuts, and one L-shape motor. Wires included in the motor set please keep nearby, we will use them later.

Connect motor to the motor plate and fix it using screws.

Same way build right motor mount.

Step 1.2 Bottom Chassis

Take bottom chassis plate and metal-ball caster wheel together with two M3x8mm screws.

Install metal-ball wheel on the bottom chassis plate.

Then use M3 nuts to fix screws.

Step 1.3 Chassis Standoffs

Take again bottom chassis plate and x4 3D-printed (or metal 55mm height) standoffs, and x8 M2.5x10mm screws.

Install screws into the 4 mounting points on the chassis bottom plate.

Install standoffs from other side of the chassis.

Make sure that standoffs are aligned with the slots on the chassis.

Step 1.4 Install Motors On The Chassis

Install motor mounts on the chassis. Insert studs located on the motor mounts into slots located on the chassis.

Step 1.5 Install Top Plate On The Robot

Take top chassis plate. It has a lot of holes to mount Intel NUCs, Nvidia Jetson (Auvidea Version), Raspberry PI, LattePanda, and AAEON UP Xtreme i11 boards. Also it has holes to mount Realsense camera in front and rear part of the robot, so you can have Visual SLAM using two Realsense cameras (that is gonna be next tutorial idea).

Place it on the standoffs by aligning motor mount studs with the rectangular slots on the top plate.

Side view.

Now fix it using four M2.5x10 mm screws.

Step 1.6 LattePanda Boards Installation

Now we're going to create a LattePanda holder. This way you can stick Velcro tape on the bottom of the plastic place and easily mount it in any place of the chassis.

Take x1 LattePanda larger plate, 4 standoffs (3D-printed or factory-made), and x4 M2.5x6mm screws. Install standoffs on the plate and fix them using screws.

Now take a LattePanda board, WiFi antennas from included in the box, and SSD M.2 disk. LattePanda 3 Delta is an x86 single board computer powered by the Intel N5105 quard-core processor. With the Intel turbo boost technology, it can reach speeds of up to 2.9 GHz. Also it has an embedded microcontroller Arduino Leonardo.

Install M.2 disk on LattePanda.

Embedded microcontroller can eliminate usage of the Raspberry board in MIKRIK project in the future, but this feature is not ready yet, I'm still working on it.

Fix antennas on the WiFi module soldered on the board and stick WiFi antennas to the plastic plate.

Finally mount board using x4 M2.5x6mm screws.

Now stick a piece of Velcro tape (soft part with loops) on the plate.

Please do the same for Raspberry PI. Mount it on the plate, and then stick Velcro tape.

Step 1.7 Stick Velcro Tape On The Chassis

Time to play with Velcro tape! Take chassis and stick two pieces of velcro tape with hooks (hard part) on the bottom plate chassis.

Here it goes.

Now stick Velcro tape on the top chassis plate.

Take Velcro and put it on the bottom side of the power bank.

Now stick a Velcro tape with loops on the 2S-battery bottom side.

Turn it over and stick Velcro tape with loops on the top side.

Step 1.8 Motor Driver

Please take DFRobot HAT Motor driver and wires that you took from motor set.

Follow the Wiring Diagram below to connect all wires on the proper pins. One motor set has two kinds of wires: 2-pin wire to power-up motor, 4-pin wire to power-up encoder and read signal encoder data.

Please follow a Raspberry PI GPIO diagram to identify the correct pin number.

Step 1.9 Wiring

After that connect wires to the correct motors. Left motor wires connect to the left motor, and its encoder. Do the same for right motor.

Connect wires

How it will look like

You might use plastic tube to do wiring more accurate.

Step 1.10 Electronics Installation On The Robot

Now install LattePanda board like that.

Next goes power bank.

After that install Li-Po batter on top.

Battery after installation.

Connect Raspberry PI with HAT Motor Driver.

And now stick Raspberry on top of the Li-Po battery.

Side view.

Step 1.11 Ethernet Connections

Now take an Ethernet cable.

And connect LattePanda and Raspberry together.

Step 1.12 Realsense Camera Mounting

Now mount D435(i) Realsense camera on the robot using 1/4-20 x 3/8" screw.

First install screw.

And then fasten camera.

Step 1.13 Power-up LattePanda Using Power Bank

Now take a USB Type C cable. One end connect to the power bank, and another to the LattePanda board.

Step 1.14 Power-up Raspberry And Motors Using Li-Po Battery

Power-up Raspberry and motors using Li-Po battery. I'm using Traxxas connector, but in the Hardware List I provided link to the T-deans battery and connector to make assembly easier and avoid soldering.

In your case you have to take Deans-T connector and cut the wire.

After that connect Deans-T connector with the same connector on the battery, like I'm doing with the Traxxas connector.

It will power-up motor driver and Raspberry. Motor driver has accept voltage in range 7-12V.

During the setup process I recommend to power-up Raspberry using an external DC power supply via USB Type C slot on Raspberry. After you successfully have system setup, you switch to the battery power supply.

Step 1.15 Final Setup

Now take wheels and install them.

Our robot part of the project is ready! We can proceed to the programming part.

Part 2 Control Computer Part Building Steps

Raspberry PI 4 Setup

Here is a ROS1 code setup that runs on Raspberry. To save time, I will not cover how to write ROS1 nodes.

Here is my Github MIKRIK repo which contains source code to run ROS1 nodes on the robot.

Step 2.1 Download Custom Raspberry Image

Please download my Ready-To-Run disk image and burn it into your microSD card. I installed on Raspberry PI Ubuntu 20.04 x64 Server and ROS1 Noetic.

Download Link Ready-To-Run MIKRIK Robot Part Disk Image

Step 2.2 Write Image File

Now take 16GB microSD card and burn image using Win32 Disk Imager on Windows or balenaEtcher on Mac. On Linux you can also do it, just use Terminal command dd

Step 2.3 Boot Raspberry

Insert microSD card into Raspberry and boot it. Connect microHDMI cable to it.

After successful boot, use a login: mikrik and password mikrik2024

Sudo password is mikrik2024

GUI will be loaded.

Step 2.4 Connect PS4 Gamepad To Raspberry

Now it is time to connect PS4 controller to it.

Enable PS4 controller into pairing mode by pressing PS button+SHARE for a 10 seconds.

Step 2.5 Pair Gamepad

Now get back to the Ubuntu Terminal and pair Raspberry with PS4 controller.

Run command to open bluetoothctl utility.

bluetoothctl

After that run

power on

And turn on scan

scan on

Find your PS4 controller in the list. It should be detected as "Wireless Controller"

Copy its address name.

Now run command to trust PS4 controller.

trust A4:AE:12:A0:C6:EF

Then pair device

pair A4:AE:12:A0:C6:EF

After that finally connect PS4 controller to Raspberry

connect A4:AE:12:A0:C6:EF

Congratulations! You connected your PS4 controller to the Raspberry. Check that status LED on the PS4 controller became a solid blue.

To test that gamepad can control robot and it works correctly, please run ROS1 nodes, following steps below.

Step 2.6 Power-up Raspberry Using Battery

Connect Li-Po battery to the motor driver and power-up Raspberry, if you didn't do that yet. It is okay if your Raspberry is connected via motor driver (battery power supply) and via Type C DC power supply at the same time.

Step 2.7 Connect Raspberry To The Wi-Fi

Connect your Raspberry to the Wi-Fi using Wi-Fi Settings panel in Ubuntu.

Then find IP by running in terminal command:

ifconfig

From this moment you no need to connect display, because now you can SSH your board from your development laptop.

Step 2.8 Launch ROS1 Nodes On MIKRIK

Open Ubuntu terminal and run ROS1 MIKRIK node

Go to the:

cd /home/mikrik/ros

Enter sudo

sudo su

Now source the environment and run ROS1 MIKRIK nodes.

source devel/setup.bash
roslaunch mikrik_description bringup.launch

After that you will see that nodes start running. roscore will be up on the Raspberry side.

Type Ctrl+C to stop program execution.

Step 2.9 Final Checks

Check that you can move motors using gamepad. See video below to review how to control robot using PS4 gamepad.

MIKRIK Robot Part setup is done!

Part 3 x86 Host Computer Setup With Robotics SDK

Host computer will control motions of the robot, and for that tutorial I'm using x86 LattePanda Delta 3. But it can be any x86 Intel CPU computer.

Host computer will read Realsense camera data and process it using Robotics SDK and Intel-proprietary ROS2 nodes for that. We will configure communication with Raspberry using ROS1-ROS2 bridge, and will install Intel Robotics SDK.

Step 3.1 Install Ubuntu 22.04 for Intel IoT

Depending on your host x86 system CPU choose an appropriate Ubuntu 22.04 version for Intel IoT to install here

If you already have installed 22.04 LTS Ubuntu Desktop, you may try not supported anymore solution after Robotics SDK 2.2 release. Replace kernel with kernel for Intel IoT. Steps are described here.

Step 3.2 Configure Ethernet

Now configure Ethernet communication with the robot part. Setup a Wired connection on your freshly-installed Ubuntu system following screenshots below.

Click on the "gear" icon on the left, and open a new window. Here check that Details tab looks the same as mine. Hardware address will be different.

Go to Identity tab. Name is doesn't matter, I put "Rpi". MAC Address choose from the drop-down list, on your system it will be different.

IPv4 Tab. Set IP address 172.100.0.100. It will be your LattePanda local IP address. Net mask 255.255.255.0. Rest in DNS section keep the same.

IPv6 tab. Disable it.

Security tab. Security disabled, default setup.

Reboot LattePanda.

Now try to ping robot part by running in LattePanda's Terminal command

ping 172.100.0.1

LattePanda must reach robot board via Ethernet.

Same way try to ping LattePanda from Raspberry side, it should also be able to reach LattePanda.

ping 172.100.0.100

Step 3.3 Setup VNC On LattePanda Ubuntu 22.04

It is pretty straightforward process that is described here.

In the end you must access your LattePanda board from your own laptop using VNC server via Wi-Fi network

Install xfce

sudo apt update
sudo apt install xfce4 xfce4-goodies

Then install vncserver

sudo apt install tigervnc-standalone-server
sudo apt install tigervnc-common
sudo apt install tigervnc-tools

Now configure password, I recommend using again mikrik2024

vncpasswd

Now update open file xstartup

vi ~/.vnc/xstartup

Update contents of it with

#!/bin/sh
unset SESSION_MANAGER
unset DBUS_SESSION_BUS_ADDRESS
exec startxfce4

Then make it executable

chmod u+x ~/.vnc/xstartup

Now get back to your laptop. Open terminal window to SSH your LattePanda board from the laptop. Windows users can install Putty to access LattePanda via SSH.

ssh amr@192.168.0.53

Then launch vncserver.

vncserver -localhost no -geometry 1920x1080 -depth 16 :42

If everything works correct, output will be like this, remember port number, I have 5942, you will have the same:

amr@robot-kit:~$ vncserver -localhost no -geometry 1920x1080 -depth 16 :42
New Xtigervnc server 'robot-kit:42 (amr)' on port 5942 for display :42.
Use xtigervncviewer -SecurityTypes VncAuth,TLSVnc -passwd /home/amr/.vnc/passwd robot-kit:42 to connect to the VNC server.

If you receive an error, first make sure that you got disconnected physical display from LattePanda board and try again. If you powered LattePanda via USB Type C and is using the same cable to connect to the display, then power the board not from a display USB C output, but for example from a USB Type C wall-adapter or laptop charger. VNC will not work if any physical display is connected, you might find fix for that on the internet.

Now try to access LattePanda VNC server from laptop.

Take your laptop, I'm using Mac, so here will be Mac setup. For Windows please dowload TightVNC Viewer app.

Open Finder. In menu choose Go -> Connect to Server.

Change my LattePanda IP Address with yours and add Port Number:

vnc://192.168.0.53:5942

Type password you set, mikrik2024.

Finally you can access LattePanda remotely! Congratulations!

Step 3.4 Setup Host Name

This way ROS1-ROS2 bridge running on LattePanda will be able to access Raspberry ROS1 instance via Ethernet.

Go to hosts file and add proper host name for the Raspberry.

sudo vi /etc/hosts

Add the line below into that file, and save it:

172.100.0.1 mxlfrbt

Try to ping Raspberry using its host name "mxlfrbt". It should work save as IP address.

ping mxlfrbt

Step 3.5 Go Back To Raspberry

Now go back to the Raspberry PI robot system terminal during this step.

Make sure that in the system it has a proper LattePanda's host name. In my case it was a "robot-kit". Depends on your LattePanda's host name in Ubuntu system, update it accordingly.

Go to:

sudo vi /etc/hosts

And update line with your system host name, my x86 computer uses name "robot-kit", yours might be different:

172.100.0.100 robot-kit

Now try to ping LattePanda from Raspberry using host name.

Run:

ping robot-kit

It should reach computer by using its hostname.

Step 3.6 Build A ROS1-ROS2 bridge

First install Docker if it is not installed on your system by following an official documentation here

ROS1-ROS2 bridge for Ubuntu 22.04 is an unofficial way based on solution from TommyChangUMD Github repo. If something is not working, check the repo's readme source, because it is updated constantly.

Clone TommyChangUMD Github repo and build docker image.

cd ~
git clone https://github.com/TommyChangUMD/ros-humble-ros1-bridge-builder.git

Now start docker image build, it will take around one hour

cd ~/ros-humble-ros1-bridge-builder
docker build . -t ros-humble-ros1-bridge-builder

After a successful build, next step is to run your new image in a container

cd ~
docker run --rm ros-humble-ros1-bridge-builder | tar xvzf -

Then you can delete builder image:

docker rmi ros-humble-ros1-bridge-builder

In case you can't execute docker commands, try to run them sudo.

If docker not installed, you can install it:

sudo apt install docker.io

To test that it works, launch ROS1 nodes from Step 2.14-2.15 on Raspberry.

Step 3.7 Launch ROS1 node on Raspberry Pi

If you didn't launch Raspberry ROS1 nodes in Step 2.8 launch them now before the bridge launch. If nodes are already running proceed to the next step.

Open Ubuntu terminal and run ROS1 MIKRIK node

Go to the:

cd mikrik/ros

Enter sudo

sudo su

Now source the environment and run ROS1 MIKRIK nodes

source devel/setup.bash
roslaunch mikrik_description bringup.launch

After that you will see that nodes start running. roscore will be up on the Raspberry side.

Step 3.9 Launch ros1_bridge on x86 robot host platform

Open script, and update IP address with you Raspberry local IP address that you set, and also Raspberry computer name.

cd mikrik_robotics_sdk/scripts
vi mikrik_bridge_launch.sh

Like this

export ROS_IP=<your-raspberry-local-ip-address-ethernet>
export ROS_MASTER_URI=http://<your-raspberry-computer-name>:11311

After you made sure that script has correct values, run the launch bridge script

cd mikrik_robotics_sdk/scripts
./mikrik_bridge_launch.sh

If everything works, there will be no errors, and in the terminal window you will see redirected topics like that

[INFO] [1717447209.481296013] [ros_bridge]: Passing message from ROS 1 std_msgs/Float64 to ROS 2 std_msgs/msg/Float64 (showing msg only once per type)
[INFO] [1717447209.481600281] [ros_bridge]: Passing message from ROS 1 std_msgs/Float64MultiArray to ROS 2 std_msgs/msg/Float64MultiArray (showing msg only once per type)
[INFO] [1717447209.482872942] [ros_bridge]: Passing message from ROS 1 sensor_msgs/Joy to ROS 2 sensor_msgs/msg/Joy (showing msg only once per type)
[INFO] [1717447209.484723585] [ros_bridge]: Passing message from ROS 1 sensor_msgs/JointState to ROS 2 sensor_msgs/msg/JointState (showing msg only once per type)
[INFO] [1717447209.587996301] [ros_bridge]: Passing message from ROS 1 tf2_msgs/TFMessage to ROS 2 tf2_msgs/msg/TFMessage (showing msg only once per type)
[INFO] [1717447209.588910535] [ros_bridge]: Passing message from ROS 1 nav_msgs/Odometry to ROS 2 nav_msgs/msg/Odometry (showing msg only once per type)
[INFO] [1717447209.589016420] [ros_bridge]: Passing message from ROS 1 tf/tfMessage to ROS 2 tf2_msgs/msg/TFMessage (showing msg only once per type)
created 2to1 bridge for topic '/rosout' with ROS 2 type 'rcl_interfaces/msg/Log' and ROS 1 type 'rosgraph_msgs/Log'

Open a second terminal window on LattePanda, and then run

source /opt/ros/humble/setup.bash
ros2 topic list

Output of the command must be like, it will print all topics redirected from Raspberry ROS1 to the ROS2 running on LattePanda

ros2 topic list
/battery
/imu
/joint_states
/joy
/mikrik/left_wheel/angle
/mikrik/left_wheel/current_velocity
/mikrik/left_wheel/pid_debug
/mikrik/left_wheel/pwm
/mikrik/left_wheel/target_velocity
/mikrik/right_wheel/angle
/mikrik/right_wheel/current_velocity
/mikrik/right_wheel/pid_debug
/mikrik/right_wheel/pwm
/mikrik/right_wheel/target_velocity
/mobile_mikrik/cmd_vel
/mobile_mikrik/odom
/parameter_events
/rosout
/tf
/tf_static

Part 4 Visual SLAM Tutorials Based On Intel Robotics SDK

To learn about Visual SLAM Mapping, Localization and Navigation on Intel Robotics SDK and Realsense camera follow my second tutorial guide on Hackster here

Deploy ROS2 Robotics SDK and VSLAM using Realsense camera

Below are some videos what it can bring to your robotics development.

Slack Support Channel

Met an issue? Please join my Slack channel or leave comments below the post.

Still To do

Wish that some of readers can contribute on the following topics.

• Add Lidar mapping, localisation and navigation using SLAM-Toolbox.
• Add Lidar and Realsense IMU readings to the CollabSLAM.

Created issues on Github.

Custom E-Learning Courses For Your Company

I'm writing a custom E-Learning courses for your company, for more details, please, contact me https://www.tmlr.pl