Weekly Robotics #385

I’m surely glad I didn’t name this newsletter Monday Robotics, because I would be in trouble for the 2nd week in a row. The reason is in the sponsored post; I’ve been working hard with the team on this. Enjoy today’s issue!

This issue is brought to you by our sponsors:

Launching Datasets and Automations [Sponsored]

Launching Datasets and Automations cover

Foxglove has launched Datasets and Automations. Datasets turn time ranges in your recordings into versioned training data, with no copies of your sensor data and direct streaming into PyTorch and Ray. Automations run AI agents on your data to triage failures as recordings arrive. To learn more about them, join our Live Demo on November 18th.


Autonomous Trail Racing with a Bicycle Robot

Autonomous Trail Racing with a Bicycle Robot cover

If you liked the robot bicycle from the RAI Institute, you have to check out this collaboration between RAI and ETH RSL. In this project, the robot bicycle navigates a bike trail with an interesting approach:

No planner hands commands to a controller. The trail is embedded in the terrain during training and reconstructed by perception at deployment.

Cool stuff! Looking forward to the paper!


I Built a Automated High Speed Camera Robot

I Built a Automated High Speed Camera Robot cover

Jeremy Fielding released this close-to-an-hour-long video about his high-speed camera robot for capturing bee shots that we last featured in the issue #368.


Notes from CoRL 2026

Notes from CoRL 2026 cover

Benjie Holson wrote a summary of CoRL conference. The Play2Perfect project Benjie mentions indeed sounds very interesting. For the summary of 2nd day, check out this article.


GR00T 2: Scaling Robot Foundation Models for Generalization, Dexterity, and Adaptation

GR00T 2: Scaling Robot Foundation Models for Generalization, Dexterity, and Adaptation cover

We introduce GR00T 2, a new generation of foundation world-action models (WAMs). GR00T 2 predicts future video and robot actions together, learning how physical interactions unfold along with the actions that produce them. The models run on parallel grippers, dexterous hands, and humanoids. Learning to predict the world, not just act in it, unlocks three new capabilities.


MOCAP4ROS2

This tool has recently reached version 2.0! Citing the website:

Motion capture for ROS 2: one interface for OptiTrack, Vicon, Qualisys and Gazebo, and tools that work with all of them — ground truth for robots, human motion analysis, recording and monitoring.

I wish I had something like this back in the day, when I was working with a MOCAP system.


CARLA 1.0 Release

CARLA 1.0 Release cover

Speaking of releases (and I’m not talking about the games coming out this week), CARLA 1.0 is out, and it bumps the Unreal Engine version, adds an agentic platform, improves sensor simulation, and adds neural reconstruction of real drives.


Our Sponsors

  • Foxglove is a purpose-built, modular platform for robotics teams to collect, organize, and learn from vast quantities of multimodal data, creating the data flywheel to safely scale from development to distributed fleets
  • Jiga connects hardware teams directly with vetted manufacturers for reliable capacity from prototype to production, combining the speed of digital manufacturing with the trust of a long-term supplier relationship
  • Point One Navigation delivers centimeter-accurate positioning to robots and autonomous systems — powered by the world’s densest fully-owned RTK corrections network.
  • LTX is a lab for building frontier open world models, developing open-weight multimodal models that understand physical space, causality, and structure.

Events

For more robotic events, check out our event page.


Want to promote your product or service in Weekly Robotics? Check out our advertising options.

issue 384