ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A answers.ros.org

ROS2-DDS-TSN integration demo by NXP

Hi ROS community!

Today we want to bring your attention to an interesting and hopefully helpful Github repository created by NXP that shows how to integrate DDS and TSN, allowing the modeling of a differential drive vehicle in the ROS 2 environment with Gazebo.

This demo proves the advantages of integrating DDS and TSN for autonomous vehicles use cases. A Gazebo environment simulates the moose test, a challenge in which a vehicle has to perform a time-critical evasive maneuver. Linking a safety-critical DDS topic to a TSN stream results in a successful completion of the moose test.

This demo uses three machines with Ubuntu 20.04, two of them are embedded ARM-based systems, connected to a TSN Ethernet switch, and a ROS 2 Foxy base which has Fast DDS pre-installed.

You can find the whole repository and the use case in the following link: GitHub - NXP/dds-tsn: Example project of DDS-TSN integration

10 Likes

Great job @Daniel_Cabezas, thank you for sharing! I’m happy to see that you used Connext DDS (and rmw_connextdds) to demonstrate the actual DDS-TSN integration :grin:

Thank you for sharing. Very interesting.

Thanks a lot, @Daniel_Cabezas, for sharing! :+1: Last week we also added a demo video to the repository, illustrating vehicle collisions caused by the network interference. Furthermore, those who are interested in the DDS-TSN integration may like our free webinar, which has another demo video about various network interference and fault handling scenarios during autonomous valet parking with Autoware.Auto. Enjoy! :slight_smile:

2 Likes

Thank you for sharing.

Maybe i missed it - do you have benchmarks on jitter/latency for TSN/Non-TSN for this example?

1 Like

Hi @AKampmann, few days ago we also published a DDS-TSN white paper, which details a similar experimental setup using Autoware.Auto Autonomous Valet Parking covered in the free DDS-TSN webinar. This white paper does include a high-level end-to-end latency and jitter analysis regarding Figure 10, which you may find interesting. When reasoning about these numbers it’s important to note that the GitHub networking setup has a single hop, while the webinar and white paper’s setup includes multiple hops. Furthermore, the software stacks are quite different. Hope this helps!

1 Like

Thank you! I will have a look at the whitepaper.