Davide Faconti retweeted
We're excited to welcome @ahcorde as the newest project lead for #ROS! He's a longtime contributor and supportive community member who enjoys helping others give back to the project.
Learn more about him here: discourse.openrobotics.org/t…
#Robotics
ALT The headshot of ahcorde. He's a smiling Spanish man with short, dark hair and beard and glasses. He's wearing a white hoodie over a white t-shirt.
I called it the last week.
Open source is experiencing the 5 stages of grief.
I choose acceptance. Users want good software, how that software was created is irrelevant
Davide Faconti retweeted
We’re excited to announce the official release of NVIDIA Isaac Lab 3.0. 🎉
Isaac Lab 3.0 brings a new modular and lightweight architecture along with major new capabilities across the robot learning workflow: multi-physics and multi-rendering support, training at scale, and streamlined deployment on real hardware.
With Newton integration, the new OVRTX renderer, RL post-training, and LEAPP deployment, there’s a lot to explore.
Here’s a look at what’s new 🧵
(1/7)
Yesterday I tried a promp:
"Analyze in dept Gazebo and find all the opportunities to improve speed and performance, keeping exactly the same identical results".
I am f****g crying. The floor has never been higher.
Davide Faconti retweeted
Fifteen years ago, doing SLAM with a single camera didn't even stand a chance.
I was working on SLAM back then. Point Cloud Library had just been released. ROS1 was picking up steam. And if you wanted good accuracy, you used a LiDAR of some sort.
Cameras alone didn't cut it. Not even stereo. Bring up a single camera and nobody took it seriously.
Then I came across this open-source model rebuilding 3D scenes from one camera, live, at around 20 frames per second, and staying stable over long sequences.
That still blows my mind.
And it isn't only perception and mapping that grew up. Position estimation came a really long way too. Factor graphs, IMU preintegration and a stack of other techniques each pushed the field forward in their own way.
If you're a robotics nerd, you know how exciting this is.
The setup that didn't stand a chance in my SLAM days is now code you can clone and run.
I will start the next week a task that 6 months ago was estimated to be 80 person weeks of work. It was never started, for that reason.
I am planning to use mostly autonomous agentic AI to do it.
I will report back the results
The values of Open Source are dying. It is the only thing about the AI revolution that makes me sad.
We have more GitHub repos than ever, but I believe we are rapidly loosing the appreciation of others work and the gratitude that existed between mainteners, contributors and users
Davide Faconti retweeted
following the discussion weeks ago, I asked codex to vibe a comparison for MPC vs RL for just table-top EE tracking, 1.6Mtok gone. It uses Pinocchio for MPC, and codes a cupy-based simulator, tuned RL to achieve better tracking accuracy (actually both errors are too small)
Davide Faconti retweeted
We're preparing a bold experiment for #EasyNav: our real Summit robot starts on one floor of our building with AMCL, drives outside with GPS, goes down a ramp, re-enters one floor below and navigates right under where it started, back on AMCL.
#NavMap
youtu.be/9Dphue8z7ZU
Davide Faconti retweeted
Super neat landing page of @tangentrobotics
Love the lil twist which visually conveys the whole point of their startup
Davide Faconti retweeted
Me casually dropping a 3,000-line PR twenty minutes after the ticket was assigned.
Davide Faconti retweeted
Argon is so good at robotics that our whole team is having an existential crisis right now, enjoy!
developers.googleblog.com/ge…
Davide Faconti retweeted
Out of curiosity, I left Claude overnight with the task of rewriting Ruckig (including the closed-source proprietary add-ons like intermediate waypoints) in modern C++.
It worked, it's slightly faster, and it let me make a 🇨🇱 reference. Wins all around.
github.com/sea-bass/charqui
In 2015, we realized that Deep Learning would replace image processing.
In the last years, Reinforcement Learning locomotion is moving the bar beyond what Model Predictive Control did.
And now AI coding is replacing fast coding itself.
We are not emotionally ready for that.
Not yet on complex system, yes on smaller self contained applications.
Ask again in one year and the answer will probably be "yes" to both
I consistently see a large number of people using AI less than I do.
The mileage I get out of it is insane, and I am shocked that others are still reluctant.
Am I getting more mileage from AI than them, or is it because I accepted that I don't need to understand the code to build a reliable system?