Stéphane Caron
@locoscaron.fosstodon.org.ap.brid.gy
I toot about open source robots and robotics.
[bridged from https://fosstodon.org/@locoscaron on the fediverse by https://fed.brid.gy/ ]
[bridged from https://fosstodon.org/@locoscaron on the fediverse by https://fed.brid.gy/ ]
The PlaCo library handles kinematic loop closures 🔄 Frame attachments can be specified in a way compatible with URDF, and a relative position task maintains loop closure during motion. Plus, the documentation is very nice: https://placo.readthedocs.io/en/v0.9.17/kinematics/loop_closures.html
November 30, 2025 at 10:32 AM
The PlaCo library handles kinematic loop closures 🔄 Frame attachments can be specified in a way compatible with URDF, and a relative position task maintains loop closure during motion. Plus, the documentation is very nice: https://placo.readthedocs.io/en/v0.9.17/kinematics/loop_closures.html
Reposted by Stéphane Caron
Training a small humanoid robot with reinforcement learning using another robot for reset.
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
September 29, 2025 at 8:48 AM
Training a small humanoid robot with reinforcement learning using another robot for reset.
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
My website (scaron.info) is currently down. I will get it back up as soon as I can, but unfortunately it won't be before next Sunday. I'm sorry for any inconvenience this may cause!
October 8, 2025 at 11:15 PM
My website (scaron.info) is currently down. I will get it back up as soon as I can, but unfortunately it won't be before next Sunday. I'm sorry for any inconvenience this may cause!
The CoRL 2025 workshop on Open-Source Hardware in the Era of Robot Learning is starting now! You can join the conversation online via live streaming: https://www.youtube.com/live/ZVPIJzF1df4
September 27, 2025 at 12:32 AM
The CoRL 2025 workshop on Open-Source Hardware in the Era of Robot Learning is starting now! You can join the conversation online via live streaming: https://www.youtube.com/live/ZVPIJzF1df4
Tomorrow, 𝗞𝗲𝗻𝘁𝗼 𝗞𝗮𝘄𝗮𝗵𝗮𝗿𝗮𝘇𝘂𝗸𝗮 will present the MEVIUS series of open-source robots: the MEVITA biped, the MEVIUS quadruped and its chimney-climbing extension KLEIYN. What makes these robots special is their use of machined metal components ordered from e-commerce. Their achieve robust locomotion […]
Original post on fosstodon.org
fosstodon.org
September 26, 2025 at 8:15 AM
Tomorrow, 𝗞𝗲𝗻𝘁𝗼 𝗞𝗮𝘄𝗮𝗵𝗮𝗿𝗮𝘇𝘂𝗸𝗮 will present the MEVIUS series of open-source robots: the MEVITA biped, the MEVIUS quadruped and its chimney-climbing extension KLEIYN. What makes these robots special is their use of machined metal components ordered from e-commerce. Their achieve robust locomotion […]
Next Saturday, 𝗧𝗲𝘁𝘀𝘂𝘆𝗮 𝗢𝗴𝗮𝘁𝗮 will present AIREC-Basic, an open robot platform designed for AI developers. His talk will cover the benefits of shifting from proprietary, closed robots to open platforms, such as extensibility and knowledge sharing. It will be accompanied by demonstrations […]
Original post on fosstodon.org
fosstodon.org
September 23, 2025 at 7:39 AM
Next Saturday, 𝗧𝗲𝘁𝘀𝘂𝘆𝗮 𝗢𝗴𝗮𝘁𝗮 will present AIREC-Basic, an open robot platform designed for AI developers. His talk will cover the benefits of shifting from proprietary, closed robots to open platforms, such as extensibility and knowledge sharing. It will be accompanied by demonstrations […]
Next Saturday, 𝗔𝗻𝘁𝗼𝗶𝗻𝗲 𝗣𝗶𝗿𝗿𝗼𝗻𝗲 will present Pollen Robotics & Hugging Face's open-source robots, including Reachy Mini, the SO-100 arm, the Amazing Hand and the Open Duck Mini. He will discuss the sim2real challenges of making the Open Duck Mini walk, and how […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
September 21, 2025 at 12:23 PM
Next Saturday, 𝗔𝗻𝘁𝗼𝗶𝗻𝗲 𝗣𝗶𝗿𝗿𝗼𝗻𝗲 will present Pollen Robotics & Hugging Face's open-source robots, including Reachy Mini, the SO-100 arm, the Amazing Hand and the Open Duck Mini. He will discuss the sim2real challenges of making the Open Duck Mini walk, and how […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
I have been trying out 𝗨𝗯𝘂𝗻𝘁𝘂 𝗧𝗼𝘂𝗰𝗵 and out of the box the system is pretty neat. Core apps are contributed by the UBports project: https://ubports.com/
With iOS and now Android becoming more locked, their clear app development workflow looks great for, say, open-source robotics projects 🙂 […]
With iOS and now Android becoming more locked, their clear app development workflow looks great for, say, open-source robotics projects 🙂 […]
Original post on fosstodon.org
fosstodon.org
September 17, 2025 at 7:51 AM
I have been trying out 𝗨𝗯𝘂𝗻𝘁𝘂 𝗧𝗼𝘂𝗰𝗵 and out of the box the system is pretty neat. Core apps are contributed by the UBports project: https://ubports.com/
With iOS and now Android becoming more locked, their clear app development workflow looks great for, say, open-source robotics projects 🙂 […]
With iOS and now Android becoming more locked, their clear app development workflow looks great for, say, open-source robotics projects 🙂 […]
On September 27, 𝗬𝘂𝗸𝗶 𝗡𝗮𝗸𝗮𝗴𝗮𝘄𝗮 will discuss risk assessment and safety design considerations for open-source hardware. She will cover hazard-disclosure practice, the roles of safety standardization and reproducibility for scientific research, and how such standards can facilitate open hardware […]
Original post on fosstodon.org
fosstodon.org
September 10, 2025 at 6:34 AM
On September 27, 𝗬𝘂𝗸𝗶 𝗡𝗮𝗸𝗮𝗴𝗮𝘄𝗮 will discuss risk assessment and safety design considerations for open-source hardware. She will cover hazard-disclosure practice, the roles of safety standardization and reproducibility for scientific research, and how such standards can facilitate open hardware […]
On September 27, 𝗚𝗿𝗲𝗴𝗼𝗶𝗿𝗲 𝗣𝗮𝘀𝘀𝗮𝘂𝗹𝘁 will present SigmaBan, an open-hardware humanoid robot that has claimed five RoboCup World Championship victories. He will recount the evolution from prototype to champion, covering the integration of ML for perception and […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
September 5, 2025 at 6:37 AM
On September 27, 𝗚𝗿𝗲𝗴𝗼𝗶𝗿𝗲 𝗣𝗮𝘀𝘀𝗮𝘂𝗹𝘁 will present SigmaBan, an open-hardware humanoid robot that has claimed five RoboCup World Championship victories. He will recount the evolution from prototype to champion, covering the integration of ML for perception and […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
On September 27, 𝗖. 𝗞𝗮𝗿𝗲𝗻 𝗟𝗶𝘂 will share the journey of creating the open-source Stanford ToddlerBot, a fully 3D-printed humanoid robot reproducible for under $6,000. She will dive into key questions on embodied data collection, sim-to-real transfer, and how […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
August 31, 2025 at 11:07 AM
On September 27, 𝗖. 𝗞𝗮𝗿𝗲𝗻 𝗟𝗶𝘂 will share the journey of creating the open-source Stanford ToddlerBot, a fully 3D-printed humanoid robot reproducible for under $6,000. She will dive into key questions on embodied data collection, sim-to-real transfer, and how […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
On September 27, 𝗞𝗲𝗻𝗻𝗲𝘁𝗵 𝗦𝗵𝗮𝘄 will present current results on democratizing dexterity with the open-source LEAP Hands. He will show how these hands can learn complex, human-like behaviors by leveraging large-scale human motion data and modern simulation […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
August 21, 2025 at 5:16 PM
On September 27, 𝗞𝗲𝗻𝗻𝗲𝘁𝗵 𝗦𝗵𝗮𝘄 will present current results on democratizing dexterity with the open-source LEAP Hands. He will show how these hands can learn complex, human-like behaviors by leveraging large-scale human motion data and modern simulation […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
Join us for the 𝗖𝗼𝗥𝗟 𝟮𝟬𝟮𝟱 𝘄𝗼𝗿𝗸𝘀𝗵𝗼𝗽 𝗼𝗻 𝗢𝗽𝗲𝗻-𝗦𝗼𝘂𝗿𝗰𝗲 𝗛𝗮𝗿𝗱𝘄𝗮𝗿𝗲 𝗶𝗻 𝘁𝗵𝗲 𝗘𝗿𝗮 𝗼𝗳 𝗥𝗼𝗯𝗼𝘁 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴! The day will bring together researchers and makers from both academia and the industry to discuss open-source robot design, integration with reinforcement learning […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
August 12, 2025 at 10:01 AM
Join us for the 𝗖𝗼𝗥𝗟 𝟮𝟬𝟮𝟱 𝘄𝗼𝗿𝗸𝘀𝗵𝗼𝗽 𝗼𝗻 𝗢𝗽𝗲𝗻-𝗦𝗼𝘂𝗿𝗰𝗲 𝗛𝗮𝗿𝗱𝘄𝗮𝗿𝗲 𝗶𝗻 𝘁𝗵𝗲 𝗘𝗿𝗮 𝗼𝗳 𝗥𝗼𝗯𝗼𝘁 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴! The day will bring together researchers and makers from both academia and the industry to discuss open-source robot design, integration with reinforcement learning […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
Here is an #Upkie wheeled biped getting up on its own: https://www.youtube.com/watch?v=ZoFQhNnQ62U A combination of reinforcement learning to find an initial trajectory, manual adjustments to transfer to the real robot, and model predictive control for balancing.
July 21, 2025 at 7:35 PM
Here is an #Upkie wheeled biped getting up on its own: https://www.youtube.com/watch?v=ZoFQhNnQ62U A combination of reinforcement learning to find an initial trajectory, manual adjustments to transfer to the real robot, and model predictive control for balancing.
Reposted by Stéphane Caron
Need for Speed or: How I Learned to Stop Worrying About Sample Efficiency
Part II of my blog series "Getting SAC to Work on a Massive Parallel Simulator" is out!
I've included everything I tried that didn't work (and why Jax PPO was different from PyTorch PPO)
araffin.github.io/post/tune-sa...
Part II of my blog series "Getting SAC to Work on a Massive Parallel Simulator" is out!
I've included everything I tried that didn't work (and why Jax PPO was different from PyTorch PPO)
araffin.github.io/post/tune-sa...
Getting SAC to Work on a Massive Parallel Simulator: Tuning for Speed (Part II) | Antonin Raffin | Homepage
This second post details how I tuned the Soft-Actor Critic (SAC) algorithm to learn as fast as PPO in the context of a massively parallel simulator (thousands of robots simulated in parallel).
araffin.github.io
July 7, 2025 at 12:11 PM
Need for Speed or: How I Learned to Stop Worrying About Sample Efficiency
Part II of my blog series "Getting SAC to Work on a Massive Parallel Simulator" is out!
I've included everything I tried that didn't work (and why Jax PPO was different from PyTorch PPO)
araffin.github.io/post/tune-sa...
Part II of my blog series "Getting SAC to Work on a Massive Parallel Simulator" is out!
I've included everything I tried that didn't work (and why Jax PPO was different from PyTorch PPO)
araffin.github.io/post/tune-sa...
Reposted by Stéphane Caron
Testing out implementing inverse dynamics in our new simulator (SIMPLE)
June 6, 2025 at 2:37 PM
Testing out implementing inverse dynamics in our new simulator (SIMPLE)
Reposted by Stéphane Caron
New in SB3 (master branch): to use n-step returns for off-policy algorithms, simply pass `n_steps=3` (it will instantiate a `NStepReplayBuffer` being the scene).
Our implementation does not use any for loops, only NumPy vectorised operations at sampling time (so the data container stays the same).
Our implementation does not use any for loops, only NumPy vectorised operations at sampling time (so the data container stays the same).
June 16, 2025 at 10:52 AM
New in SB3 (master branch): to use n-step returns for off-policy algorithms, simply pass `n_steps=3` (it will instantiate a `NStepReplayBuffer` being the scene).
Our implementation does not use any for loops, only NumPy vectorised operations at sampling time (so the data container stays the same).
Our implementation does not use any for loops, only NumPy vectorised operations at sampling time (so the data container stays the same).
This week I was reminded that nonlinear car-like models are locally controllable, but not their linearization. For instance, the nonlinear car-like model with n trailers is controllable (Robot Motion Planning and Control, p. 77) but its linearization is not (same book, p. 176).
May 18, 2025 at 9:42 AM
This week I was reminded that nonlinear car-like models are locally controllable, but not their linearization. For instance, the nonlinear car-like model with n trailers is controllable (Robot Motion Planning and Control, p. 77) but its linearization is not (same book, p. 176).
I'm trying out Pixi for the CI of Pink. I appreciate a lot how `pixi run` does locally what will run in GitHub Actions, with the magic of conda-forge to get all dependencies and environments right.
May 7, 2025 at 4:48 PM
I'm trying out Pixi for the CI of Pink. I appreciate a lot how `pixi run` does locally what will run in GitHub Actions, with the magic of conda-forge to get all dependencies and environments right.
A Lego reaction wheel pendulum that illustrates palatably what state estimation and balancing by PID control look like: https://www.youtube.com/watch?v=WObG2LoSEwQ
Note the point where modules get coupled (e.g. PID gains and target variation rate) and there are competing ways to address the […]
Note the point where modules get coupled (e.g. PID gains and target variation rate) and there are competing ways to address the […]
Original post on fosstodon.org
fosstodon.org
February 1, 2025 at 8:49 AM
A Lego reaction wheel pendulum that illustrates palatably what state estimation and balancing by PID control look like: https://www.youtube.com/watch?v=WObG2LoSEwQ
Note the point where modules get coupled (e.g. PID gains and target variation rate) and there are competing ways to address the […]
Note the point where modules get coupled (e.g. PID gains and target variation rate) and there are competing ways to address the […]
One way to balance on wheels by PID control is to cascade two loops: an inner loop commanding wheel velocities from the body tilt error, and an outer loop commanding body tilt from a target ground velocity: https://www.youtube.com/watch?v=k_vYA-xRXyo
How This R/C Car Can go STUPID Fast While Balancing on 2 Wheels
Please consider SUBSCRIBING and Supporting me on Patreon! https://www.patreon.com/GearDownForWhatLearn More about ODrive: https://odriverobotics.comGet The F...
www.youtube.com
January 11, 2025 at 10:24 AM
One way to balance on wheels by PID control is to cascade two loops: an inner loop commanding wheel velocities from the body tilt error, and an outer loop commanding body tilt from a target ground velocity: https://www.youtube.com/watch?v=k_vYA-xRXyo
I'm adding a test set of differential IK problems to qpbenchmark: https://github.com/qpsolvers/ik_qpbenchmark
December 18, 2024 at 4:51 PM
I'm adding a test set of differential IK problems to qpbenchmark: https://github.com/qpsolvers/ik_qpbenchmark
New video: model predictive control running on an #Upkie wheeled biped for going around outdoors: https://www.youtube.com/watch?v=pZhLYKkwM1o The shot was thoroughly unplanned, overlays go wherever there was space left 😁
Model predictive control on Upkie wheeled bipeds
Model predictive control running on an #Upkie wheeled biped for balancing and going around at roughly 1.5 m/s. Upkies run on fully open-source software and h...
www.youtube.com
December 12, 2024 at 6:05 PM
New video: model predictive control running on an #Upkie wheeled biped for going around outdoors: https://www.youtube.com/watch?v=pZhLYKkwM1o The shot was thoroughly unplanned, overlays go wherever there was space left 😁