Paweł Budzianowski (@pfbudzianowski)

로봇 데이터 수집 확장은 까다롭다는 내용의 안내로, 텔레오퍼레이터마다 조작 방식이 다르고 장비마다 특성이 있어 잘못된 에피소드가 정책 학습에 악영향을 줄 수 있다고 지적합니다. 이를 해결하기 위해 유효한 예시는 유지하면서 노이즈를 걸러내는 품질 필터 개선 기법을 공유한다고 발표한 기술 안내 게시물입니다.

https://x.com/pfbudzianowski/status/2031135588498247922

#robotics #dataquality #dataset #robotlearning

Paweł Budzianowski (@pfbudzianowski) on X

Scaling robot data collection is messy. Every teleoperator moves differently, every rig has quirks, and bad episodes silently poison your policies. We share how we improve quality filters that catch the noise without throwing away good examples.

X (formerly Twitter)

Jim Fan (@DrJimFan)

새 이정표: 세계 모델(world model) 백본으로 로봇 파운데이션 모델을 학습시켜 제로샷 오픈월드 프롬프팅(새로운 동사·명사·환경에 대한 즉시 대응) 능력을 부여했다고 발표. 세계 모델이 픽셀로 올바른 미래를 '꿈'꾸면 로봇이 모터로 잘 실행할 수 있다는 개념 제시.

https://x.com/DrJimFan/status/2019112603637920237

#robotics #foundationmodel #worldmodel #zeroshot #robotlearning

X

X (formerly Twitter)

The Humanoid Hub (@TheHumanoidHub)

Robbyant가 LingBot-VLA를 발표했습니다. 오픈소스 Vision-Language-Action(VLA) 모델로 약 2만 시간의 실제 듀얼암 로봇 데이터로 사전학습되었고, 9개 서로 다른 로봇 체계(embodiments)에서 강한 일반화 성능을 보이며 데이터 증가에 따라 일관되게 성능이 향상된다고 합니다. 또한 π₀.₅, GR00T N1.6 등을 능가한다고 주장합니다.

https://x.com/TheHumanoidHub/status/2017337216054575513

#visionlanguageaction #robotics #opensource #robotlearning

X

X (formerly Twitter)
I see some startups in #robotlearning continuously releasing new VLA models. I wonder how much of them provide faked benchmarks and fake teleoped demo videos.
Humanoid robots are officially teaching themselves.
Not following code. Not repeating routines.
They’re figuring things out — in real time.
Boston Dynamics’ Atlas now runs on one unified AI brain that controls its walking, bending, and grabbing. It can even pick up dropped objects—without ever being explicitly programmed to.
This is called emergent behavior.
We’ve seen it with language models… now it’s happening with motion.
And it’s not just Atlas.
#HumanoidRobots #AIInRobotics #RobotLearning

🚀 Join us for a discussion on the latest advancements in robot learning with Oliver Hausdörfer, a PhD student from TUM. We'll cover topics such as robotics, drones, and other real-world applications. @andreasofthings
#robotlearning #AI #automation #realworld #robotics #drones #live

Click here to join the live stream:
https://lnkd.in/eyD2_W5K

LinkedIn

This link will take you to a page that’s not on LinkedIn

🗺 CoRoboLearn is jointly organized with:
- Loris Roveda, Associate Professor @ Politecnico di Milano, Italy & Senior Researcher @ IDSIA USI-SUPSI, Switzerland
- Asad A. Shahid, Researcher @ IDSIA USI-SUPSI, Switzerland
- Yue Hu, Assistant Professor @ University of Waterloo, Canada
- Cheng Tang, Research Assistant @ University of Waterloo, Canada

📲 We are looking forward to meeting you in Munich, stay tuned for details!

#CoRL2024 #Robotics #MachineLearning #RobotLearning #Research #Workshop

Stanford’s HumanPlus Robot Learns Skills by Watching Humans: A New Era in Robotics

Stanford researchers develop HumanPlus, a humanoid robot that mimics human movements and learns new skills by observation. Discover how this innovative robot can perform tasks like playing piano and ping-pong.

Tech Chill

Self-driving cars, logistics, manufacturing, home automation: Autonomous robots are becoming increasingly important for our daily lives. Jan Peters, Director of our program 'ELLIS Robot Learning: Closing the Reality Gap!', gives insights into this field & tips for young scientists seeking a career in this area.

👉 https://ellis.eu/news/ellis-program-robot-learning-closing-the-reality-gap-an-interview-with-one-of-the-directors

#robotlearning #ELLISProgram

ELLIS Program ‘Robot Learning: Closing the Reality Gap!’ - An interview with one of the Directors

The ELLIS mission is to create a diverse European network that promotes research excellence and advances breakthroughs in AI, as well as a pan-European PhD program to educate the next generation of AI researchers. ELLIS also aims to boost economic growth in Europe by leveraging AI technologies.

European Lab for Learning & Intelligent Systems

Very excited to announce that Leila Takayama (UCSC) is joining our #RobotLearning workshop at #NeurIPS2022 to talk about:
'Why Robots Need Social Skills'

Please find the video under
https://youtu.be/RWzqGY6jiq4

And join us on the 9.12!

Why Robots Need Social Skills - Leila Takayama

YouTube