Robotic systems present an embedded computing challenge, as they require real-time communication with low latency, multi-sensor integration, and deterministic performance, all as part of a complex middleware stack. ROS 2 (Robot Operating System 2) has become a de-facto standard for robotics due to its Data Distribution Service (DDS) communication architecture, modular package ecosystem, and support for embedded Linux.

This blog post documents our technical exploration of bringing a ROS 2-based robotic prototype onto Toradex’s Verdin iMX8M Plus System on Module, a platform well-suited for robotics due to its NPU, heterogeneous cores, and camera interfaces.

The goal of this post is not to serve as a step-by-step robotics tutorial, but rather to document our engineering evaluation of ROS 2 workflows on Toradex hardware. We highlight what worked, what didn’t, and how Yocto-based builds compare to Torizon’s containerized approach, in terms of speed, maintainability, and deployment. We also provide example commands so readers can reproduce key steps.

ROS 2 is not an operating system in the traditional sense; rather, it is an open-source software framework that helps developers build robotic applications. You can think of it this way: Linux is the processor’s OS, while ROS 2 is the OS of the actual robot.

A reason why ROS 2 is widely used in robotics is because of its support for real-time communication. This is enabled by DDS, which allows developers to tune how messages are delivered depending on the needs of each task. ROS 2 also benefits from a large ecosystem of ready-to-use packages for navigation, perception, visualization, and hardware drivers.

For our project, we chose ROS 2 Humble because it offers long-term support and runs well on ARM-based embedded hardware like the Toradex Verdin iMX8M Plus. Using ROS 2 made it easier to organize the robot’s software, integrate multiple sensors, and reuse existing tools while still having the flexibility to build our own features.

• Yocto Scarthgap 5.0 with the meta-ros layer.
• ROS 2 Humble distribution.
• A range of onboard sensors:
  • DHT20 (temperature/humidity)
  • Ambient light sensor
  • IMU (accelerometer, gyroscope, magnetometer)
  • Ultrasonic range finders
  • 8 m indoor LiDAR
  • USB camera

We packaged the application in a Debian-style build on an Ubuntu base container, which enabled quick iteration. The robot ran successfully on the Toradex Dahlia and Mallow carrier board with the Verdin iMX8M Plus System on Module.

• Kernel patching & driver integration were needed for custom sensors.
• Maintaining custom Yocto layers required significant overhead.
• System updates had to be manually managed.

Although we achieved functional results, it became clear we were duplicating tooling that Toradex already provides through Torizon OS.

Torizon OS is Toradex’s open-source industrial Linux platform, built on Yocto but designed for ease of development and maintenance.

1. Stable Base OS
   • Long-term support (LTS) with regular security patches.
   • No need to maintain custom Yocto builds.
2. Containerized Deployment
   • Encapsulated ROS 2 stack in Docker containers.
   • Eliminated conflicts with the base OS.
   • Identical environments across development and deployment devices.
3. Developer-Friendly Tooling
   • Torizon IDE extension for Visual Studio Code.
   • Remote debugging, cross-compilation, and CI/CD integration.
4. Future-Proof with OTA Updates
   • Built-in Over-The-Air (OTA) updates for application and system.
   • Secure boot and SBOM generation for compliance.

With this, iteration cycles shrank drastically. Instead of rebuilding entire Yocto images, we now develop inside an Ubuntu-based ROS container and deploy via Torizon in minutes.

1. Prerequisites
   • Toradex board (e.g., Verdin iMX8MP SOM on Mallow carrier board).
   • Toradex Easy Installer flashed with Torizon OS.
   • Host PC running Ubuntu 22.04 with ROS 2 Humble.
2. Flash Torizon OS
   • Use Toradex Easy Installer to flash Torizon OS from the online server.
   • Boot the board and open the debug console.
   • Login:
     • Username: torizon
     • Password: torizon
3. Running a Prebuilt ROS 2 Package
   
   # Step 1: Pull Sibrain’s prebuilt ROS 2 base image

   docker pull sibrain/rosbase

   # Step 2: Run container with host networking

   docker run --network host -it sibrain/rosbase:latest

   # Step 3: Setup ROS environment

   source /opt/ros/humble/setup.bash 

   # Step 4: Run demo talker node

   ros2 run demo_nodes_py talker

   From your host PC (on the same network):

   ros2 topic list  
   ros2 topic echo /chatter
   

4. Running a Custom ROS 2 Package
   We use two containers:
   • rosbase → ROS setup container
   • hello_ros → Custom package container
   # Step 1: Pull required images

   docker pull sibrain/rosbase
   docker pull sibrain/hello_ros  
   

   # Step 2: Start containers

   docker run -d --name rosbase_running --network host -v ros_opt:/opt -it sibrain/rosbase:latest  
   docker run -d --name hello_ros --network host -v ros_opt:/opt -it sibrain/hello_ros  
   

   # Step 3: Build custom ROS package

   docker exec -it hello_ros bash  
   cd src/  
   source /opt/ros/humble/setup.bash  
   colcon build  
   source install/setup.bash  
   

   # Step 4: Run custom node

   ros2 run ros_py_pkg py_node  
   From your host PC:
   ros2 topic list  
   ros2 topic echo /helloros  
   

This workflow enables rapid testing of both generic and application-specific ROS packages on Toradex hardware.

• Publish ROS 2 containers in the Torizon Demo Gallery.
• Provide a ROS 2 template for the Torizon IDE.
• Showcase upcoming demos at trade events, including a ROS 2 drone prototype.

With ROS 2 + Torizon on Toradex hardware, you can focus on robot intelligence and innovation instead of infrastructure maintenance.

We also recommend reading the How Torizon and ROS 2 Empower your Robotics Projects blog post to get a deeper look at how you can start your ROS 2 project with Torizon and SiBrain’s blog post on their step by step development process for the robot prototype.

Author:

Leave a comment