Teams,
first of all we’d like to apologize for the radio silence during last two weeks. We’ve took our time, analyzed the messages you’ve sent us and tested the issues ourselves. During that time we’ve learnt that many changes are justified. This message is meant to be a summary of them in a version more readable than a simple git diff
.
Date and team limit changes
2025.07.15 - 20 Teams - ROS Test - We’ve decided to add this step as we’ve not managed to test the ROS connectivity last time. Gist will be similar to what was meant for the Connectivity Test (regarding ROS) but there will be an additional step involving on-site compute. On-site compute will be available this time.
2025.07.29 - 10 Teams - Payload Test
2025.08.12 - 10 Teams - Challenge Test
2025.08.26 - 5 Teams - Finals - First Challenge
Challenge 2 drop
With great regret we decided to skip the Challenge 2 this year entirely. We want to prioritize quality over quantity and considering the amount of details that matter - we decided that Challenge 1 will present enough problems to tackle for both us and the Teams.
The actual competition time slot lengths remain the same, but we will have less days during the Finals.
Challenge 1 configuration change
The goals of the Challenge remain the same, however however both hardware and software stack will change.
Hardware-wise - we initially were under impression that the 4-camera setup from Stereolabs is production-ready in general. While it may be in some cases, but compatibility with ROS2 Jazzy requires some extra steps that we would not like to add to the Teams’ agenda. This is why we will be changing the setup to a slightly more familiar lidar + 4xRGB cameras. Both elements will have ROS nodes prepared (and operated) by us.
Software-wise - ROS2 Jazzy turned out not to be the problem in the big picture (there were a couple of compatibility issues here and there but no deal-breakers) - so it stays. We’ve simplified the network stack though. On the on-site LAN everything speaks plain Cyclone DDS with multicast discovery — no Discovery Server to worry about. Only the ROS 2 Router container (which bridges the LAN to your laptop over Husarnet) uses Fast DDS plus its Discovery Server, because that combo still performs best across the internet. So: inside the venue run Cyclone; when you connect remotely, point your nodes to the router’s superclient.xml
profile and use Fast DDS.
On-site Compute
We’ve managed to secure some computing power to be used on-site. Each team will be assigned a Virtual Machine with come CPU resources (that we intend to be used mostly for proxying data and the navigation stack) but with no real GPU power - so be ready to setup a forwarding of some data (i.e. video) to your own sites - with decent image analysis power.
Access to the on-site compute resources will be given one week before the ROS Test (2025.07.08).
Technical details
As usual - technical details can be found in Technical Handbook - which we’ve updated to reflect the latest changes.
Closing remarks
We’re sorry for introduction of such fundamental changes while the Competition is in progress, but we do believe that it’s in the best interest of the quality of the whole endavor.