Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Excited to introduce PyRoki ("Python Robot Kinematics"): easier IK, trajectory optimization, motion retargeting... with an open-source toolkit on both CPU and GPU

ChungMinKim's profile picture

Chung Min Kim

1,155 subscribers

117,384 Aufrufe • vor 1 Jahr •via X (Twitter)

Wissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

27 Kommentare

Profilbild von Chung Min Kim
Chung Min Kimvor 1 Jahr

PyRoki's focus is flexibility, but it's also fast: up to 1.4-1.7x faster than cuRobo for GPU-accelerated IK! (with higher success and lower errors, see paper ✨)

Profilbild von Chung Min Kim
Chung Min Kimvor 1 Jahr

This work wouldn't have been possible without my collaborators @brenthyi @redstone_hong @YiMaTweets @Ken_Goldberg @akanazawa ! The toolkit is available here: Website: Code:

Profilbild von Kevin Zakka
Kevin Zakkavor 1 Jahr

Happy to see this public, congrats @ChungMinKim !

Profilbild von Chung Min Kim
Chung Min Kimvor 1 Jahr

Thank you!! Excited to have this out :-)

Profilbild von Max Fu
Max Fuvor 1 Jahr

Congrats on the release!

Profilbild von Vector Wang
Vector Wangvor 1 Jahr

Can it do ik on a 5dof arm? Like LeRobot so100?

Profilbild von Chung Min Kim
Chung Min Kimvor 1 Jahr

:-)

Profilbild von Vector Wang
Vector Wangvor 1 Jahr

Definitely try it this week and use it for my XLeRobot!

Profilbild von Shreyas Kapur
Shreyas Kapurvor 1 Jahr

Incredible work ✨ This is so cool!!

Profilbild von Chung Min Kim
Chung Min Kimvor 1 Jahr

🥹 thank u 👉👈

Profilbild von Danijar Hafner
Danijar Hafnervor 1 Jahr

Looks cool, congrats!!

Profilbild von Amy Lu
Amy Luvor 1 Jahr

This is so cool!! Yay Chung Min 🤩🌟

Profilbild von Michael Cho - Rbt/Acc
Michael Cho - Rbt/Accvor 1 Jahr

Great stuff! We can always use better IK 😅

Profilbild von jessica dai
jessica daivor 1 Jahr

yay!!!!

Profilbild von Carlos DP 🤖🇺🇸
Carlos DP 🤖🇺🇸vor 1 Jahr

You guys are killing it today, damn lol

Profilbild von RobotSoul
RobotSoulvor 1 Jahr

Need more regularization?

Profilbild von Youngsun Wi ✈️ RSS 2026
Youngsun Wi ✈️ RSS 2026vor 1 Jahr

Hey! amazing work👍 excited to try it out

Profilbild von Reborn
Rebornvor 1 Jahr

This looks awesome, love the clean IK and retargeting workflow.

Profilbild von Daily Digital Digest
Daily Digital Digestvor 1 Jahr

Cooool!

Profilbild von CyberNetics
CyberNeticsvor 1 Jahr

Hey, PyRoki sounds cool, but how does it stack up against other tools for robot kinematics? What makes its IK and trajectory optimization easier to use?

Profilbild von Anon
Anonvor 1 Jahr

No quadrupeds? 😔

Profilbild von ObotX - Open Source Robotics
ObotX - Open Source Roboticsvor 1 Jahr

Wow! Awesome! Does it work also with closed loop IK like: ? It might be the future of robotics as its much simpler, less expensive with high load capacity

Profilbild von Raj
Rajvor 1 Jahr

This is pretty cool - will check it out. QQ: Did you find PyRoki perform better in collisions also than cuRobo?

Profilbild von BlockChan
BlockChanvor 10 Monaten

wen this gonna be public product

Profilbild von 51WORLD
51WORLDvor 1 Jahr

Cool project👏More GPU-friendly robotics tools are exactly what's needed

Profilbild von Amit Ika
Amit Ikavor 1 Jahr

@TheHumanoidHub Can it consider joint angle limits when calculating IK solutions? Or collision avoidance with known objects in the workspace?

Profilbild von ScottLabs
ScottLabsvor 1 Jahr

Amazing work! I knew Python was versatile but never even considered it for robot simulations. Love it!

Ähnliche Videos

1.7× faster inverse kinematics in pure Python, GPU-accelerated! Fully open-source and built for scale [📍Bookmark for later] PyRoki, a modular toolkit for robot kinematic optimization, supporting - inverse kinematics - trajectory optimization, and - motion retargeting. Built for flexibility, but also fast: ✅ Runs on CPU, GPU, and TPU ✅ Outperforms cuRobo in speed, success rate, and accuracy ✅ Fully in Python, easy to integrate and extend Whether you’re working with industrial arms, simulation environments, or humanoids, PyRoki is designed to fit your stack and scale with your hardware. Thank you, Chung Min Kim, for sharing!! Website: Code: Paper: This one’s for researchers, builders, and anyone tired of debugging their IK stack.
0:49

1.7× faster inverse kinematics in pure Python, GPU-accelerated! Fully open-source and built for scale [📍Bookmark for later] PyRoki, a modular toolkit for robot kinematic optimization, supporting - inverse kinematics - trajectory optimization, and - motion retargeting. Built for flexibility, but also fast: ✅ Runs on CPU, GPU, and TPU ✅ Outperforms cuRobo in speed, success rate, and accuracy ✅ Fully in Python, easy to integrate and extend Whether you’re working with industrial arms, simulation environments, or humanoids, PyRoki is designed to fit your stack and scale with your hardware. Thank you, Chung Min Kim, for sharing!! Website: Code: Paper: This one’s for researchers, builders, and anyone tired of debugging their IK stack.

Ilir Aliu

18,641 Aufrufe • vor 1 Jahr

Excited to open-source GMR: General Motion Retargeting. Real-time human-to-humanoid retargeting on your laptop. Supports diverse motion formats & robots. Unlock whole-body humanoid teleoperation (e.g., TWIST). video with 🔊
0:32

Excited to open-source GMR: General Motion Retargeting. Real-time human-to-humanoid retargeting on your laptop. Supports diverse motion formats & robots. Unlock whole-body humanoid teleoperation (e.g., TWIST). video with 🔊

Yanjie Ze

88,323 Aufrufe • vor 1 Jahr

I’m excited to introduce MASTER OSINT v2, a powerful Python toolkit for cybersecurity enthusiasts and investigators. This updated version includes 12 modules that streamline the collection and analysis of open-source intelligence. #OSINT #Cybersecurity
9:58

I’m excited to introduce MASTER OSINT v2, a powerful Python toolkit for cybersecurity enthusiasts and investigators. This updated version includes 12 modules that streamline the collection and analysis of open-source intelligence. #OSINT #Cybersecurity

D4rk_Intel

11,795 Aufrufe • vor 1 Jahr

We’re excited to introduce Pocket TTS: a 100M-parameter text-to-speech model with high-quality voice cloning that runs on your laptop—no GPU required. Open-source, lightweight, and incredibly fast. 🧵👇
1:44

We’re excited to introduce Pocket TTS: a 100M-parameter text-to-speech model with high-quality voice cloning that runs on your laptop—no GPU required. Open-source, lightweight, and incredibly fast. 🧵👇

kyutai

238,618 Aufrufe • vor 8 Monaten

Excited to share with Inria Flowers team ADTool, an open-source Python toolkit using curiosity-driven AI to explore behavior diversity of complex systems (cellular automata, physics models, ...)!🐍 Serving off-the-shelf strategies like IMGEP You said IMGEP? 👇
2:03

Excited to share with Inria Flowers team ADTool, an open-source Python toolkit using curiosity-driven AI to explore behavior diversity of complex systems (cellular automata, physics models, ...)!🐍 Serving off-the-shelf strategies like IMGEP You said IMGEP? 👇

Inria Flowers team

13,752 Aufrufe • vor 2 Jahren

Video2Humanoid. Still trouble with bad retargeted humanoid motions? Humanoids now are Easy to track any video using our new Neural Motion Retargeting (NMR) Method. Optimization-based motion retargeting methods like IK and GMR are solving a non-convex problem frame by frame, which makes them sensitive to initialization, hard to tune, and prone to poor local minima. The result is familiar: joint discontinuities, self-collisions, and unstable foot contact. Our solution is simple but effective : instead of optimizing each frame independently, learn the mapping between human motion and robot motion as a distribution via networks. 📊Results on Unitree G1 - 0 joint jumps - 54% fewer self-collision frames - 61% fewer joint limit violations - Faster convergence for downstream control policy training NMR turns motion retargeting from a fragile optimization problem into a learned, scalable pipeline for more stable humanoid motion. please visit for more visualization results. opensource:
0:54

Video2Humanoid. Still trouble with bad retargeted humanoid motions? Humanoids now are Easy to track any video using our new Neural Motion Retargeting (NMR) Method. Optimization-based motion retargeting methods like IK and GMR are solving a non-convex problem frame by frame, which makes them sensitive to initialization, hard to tune, and prone to poor local minima. The result is familiar: joint discontinuities, self-collisions, and unstable foot contact. Our solution is simple but effective : instead of optimizing each frame independently, learn the mapping between human motion and robot motion as a distribution via networks. 📊Results on Unitree G1 - 0 joint jumps - 54% fewer self-collision frames - 61% fewer joint limit violations - Faster convergence for downstream control policy training NMR turns motion retargeting from a fragile optimization problem into a learned, scalable pipeline for more stable humanoid motion. please visit for more visualization results. opensource:

Xiao-Xiao Long

24,245 Aufrufe • vor 5 Monaten

Excited to share Toolkit, a project I built for Theo - t3.gg’s T3 Cloneathon Mix-and-match GitHub data, Notion, Memory, Code Interpreter, and much more through an interactive chat Fully open source and looking for more Toolkit developers! Details below…
1:44

Excited to share Toolkit, a project I built for Theo - t3.gg’s T3 Cloneathon Mix-and-match GitHub data, Notion, Memory, Code Interpreter, and much more through an interactive chat Fully open source and looking for more Toolkit developers! Details below…

Jason Hedman

119,602 Aufrufe • vor 1 Jahr

Today, we’re excited to announce and launch Reflex Cloud on Product Hunt! Reflex is an open-source framework for building and deploying data and AI web apps in pure Python. Frontend and Backend in Pure Python: No JavaScript required! With Reflex Cloud, you can now deploy, manage, and scale your Python apps with just a single command! If you're a Python developer, an upvote or a share would mean a lot to us :)💪❤️
3:04

Today, we’re excited to announce and launch Reflex Cloud on Product Hunt! Reflex is an open-source framework for building and deploying data and AI web apps in pure Python. Frontend and Backend in Pure Python: No JavaScript required! With Reflex Cloud, you can now deploy, manage, and scale your Python apps with just a single command! If you're a Python developer, an upvote or a share would mean a lot to us :)💪❤️

Reflex

55,100 Aufrufe • vor 1 Jahr

🤩Excited to share SANA-WM: a 2.6B open-source world model for minute-scale 720p video generation. Given one image + text + a 6-DoF camera trajectory, it synthesizes action-controllable 60s worlds on a single GPU. Project: Paper:
1:14

🤩Excited to share SANA-WM: a 2.6B open-source world model for minute-scale 720p video generation. Given one image + text + a 6-DoF camera trajectory, it synthesizes action-controllable 60s worlds on a single GPU. Project: Paper:

Haoyi Zhu

124,608 Aufrufe • vor 4 Monaten

At Advancing AI 2026, AMD went after the humanoid "robot brain," directly challenging NVIDIA's Jetson Thor at the edge. Its new Kria AI Robotics Developer Platform is an open, turnkey system putting CPU, GPU, NPU and FPGA on one board. The X100 chip is the robot's "brain," while dedicated FPGAs control the joints, motion, and sensors in real time, something AMD has long done well. AMD claims the Kria/X100 robot module beats NVIDIA's Jetson Thor T5000 by up to 3.4× on real-time reaction, 2.3× on concurrent agents, and 1.6× on CPU (AMD's own numbers). Here's Lisa Su meeting Generative Bionics' humanoid. AMD is both an investor in and tech partner to the Italian startup.
0:33

At Advancing AI 2026, AMD went after the humanoid "robot brain," directly challenging NVIDIA's Jetson Thor at the edge. Its new Kria AI Robotics Developer Platform is an open, turnkey system putting CPU, GPU, NPU and FPGA on one board. The X100 chip is the robot's "brain," while dedicated FPGAs control the joints, motion, and sensors in real time, something AMD has long done well. AMD claims the Kria/X100 robot module beats NVIDIA's Jetson Thor T5000 by up to 3.4× on real-time reaction, 2.3× on concurrent agents, and 1.6× on CPU (AMD's own numbers). Here's Lisa Su meeting Generative Bionics' humanoid. AMD is both an investor in and tech partner to the Italian startup.

The Humanoid Hub

37,885 Aufrufe • vor 2 Monaten

Unitree Humanoid Robots Updated With Open-Source Full-Body Motion Dataset 🤗 We’ve adapted motion capture data for Unitree H1, H1-2, and G1 humanoid robots, using numerical optimization methods to make their movement trajectories more natural. Open-Source： #AI #OpenSourceDataset #Unitree #EmbodiedAI #Humanoid #DataCollection #AGI #Dance
1:04

Unitree Humanoid Robots Updated With Open-Source Full-Body Motion Dataset 🤗 We’ve adapted motion capture data for Unitree H1, H1-2, and G1 humanoid robots, using numerical optimization methods to make their movement trajectories more natural. Open-Source： #AI #OpenSourceDataset #Unitree #EmbodiedAI #Humanoid #DataCollection #AGI #Dance

Unitree

198,619 Aufrufe • vor 1 Jahr

SwissGL Demo: Particle Lenia with alien SOUNDS! Audio is fully rendered on GPU with GLSL shaders and streamed to WebAudio thanks to the new async GPU->CPU read feature. LIVE DEMO: Press "toggleAudio" in settings for sound 4K YouTube version: Source code:
0:28

SwissGL Demo: Particle Lenia with alien SOUNDS! Audio is fully rendered on GPU with GLSL shaders and streamed to WebAudio thanks to the new async GPU->CPU read feature. LIVE DEMO: Press "toggleAudio" in settings for sound 4K YouTube version: Source code:

Alex Mordvintsev

34,580 Aufrufe • vor 3 Jahren

Spatial optimization has applications in nearly every field but it's underutilized. In 2026, I'm working on open source and lightning-fast tools to make spatial optimization accessible. Shown: interactively solving a "location-allocation" problem in an app with instant results.
0:43

Spatial optimization has applications in nearly every field but it's underutilized. In 2026, I'm working on open source and lightning-fast tools to make spatial optimization accessible. Shown: interactively solving a "location-allocation" problem in an app with instant results.

Kyle Walker

28,495 Aufrufe • vor 9 Monaten

Just updated #raylib animation system to support a more flexible and easy-to-use blending feature, with CPU or GPU skinning. Several code examples added! 🔥 Try it on web: Source code:
0:44

Just updated #raylib animation system to support a more flexible and easy-to-use blending feature, with CPU or GPU skinning. Several code examples added! 🔥 Try it on web: Source code:

Ray

17,202 Aufrufe • vor 7 Monaten

HelixDB () makes RAG and AI retrieval easier to build and manage with an ultra-performant, open source graph-vector database. Congrats on the launch, Lam Sambert and Xav! Check it out on GitHub:
0:42

HelixDB () makes RAG and AI retrieval easier to build and manage with an ultra-performant, open source graph-vector database. Congrats on the launch, Lam Sambert and Xav! Check it out on GitHub:

Y Combinator

67,993 Aufrufe • vor 1 Jahr

Democrats Introduce Bill to Require Age Verification on Linux, Windows "Age Attestation on Computing Devices" (Colorado State Bill 26-051) would require age verification on all Operating Systems (both Open Source and proprietary), with fines for violations.
13:03

Democrats Introduce Bill to Require Age Verification on Linux, Windows "Age Attestation on Computing Devices" (Colorado State Bill 26-051) would require age verification on all Operating Systems (both Open Source and proprietary), with fines for violations.

The Lunduke Journal

349,226 Aufrufe • vor 7 Monaten

X Square Robot has introduced TwinDEX: a co-designed pair of three-finger, 9-DoF dexterous hands, one wearable for data collection, one matched robot end effector for deployment. The wearable and the robot hand share the same kinematics, geometry, materials, look, and sensors, so human demos map straight to the robot without retargeting. Human data collection improves throughput 5.3x compared to teleoperation, and a policy can be trained with a few hundred episodes without any robot data.
3:51

X Square Robot has introduced TwinDEX: a co-designed pair of three-finger, 9-DoF dexterous hands, one wearable for data collection, one matched robot end effector for deployment. The wearable and the robot hand share the same kinematics, geometry, materials, look, and sensors, so human demos map straight to the robot without retargeting. Human data collection improves throughput 5.3x compared to teleoperation, and a policy can be trained with a few hundred episodes without any robot data.

The Humanoid Hub

277,022 Aufrufe • vor 1 Monat