fly51fly (@fly51fly)

인간과 유사한 조작을 목표로 한 연구가 arXiv에 공개되었습니다: 강화학습(RL)로 보강한 원격조작(teleoperation)과 'Mixture-of-Dexterous-Experts VLA' 방식을 제안합니다. 저자 T Tang, X Ji, W Xing, C Hao 등(Shanghai Jiao Tong University & Sharpa)이 공동 발표한 논문(2603.08122)입니다. 인간 수준의 섬세한 조작 향상을 목적으로 합니다.

https://x.com/fly51fly/status/2033297975972417595

#robotics #teleoperation #reinforcementlearning #manipulation

fly51fly (@fly51fly) on X

[RO] Towards Human-Like Manipulation through RL-Augmented Teleoperation and Mixture-of-Dexterous-Experts VLA T Tang, X Ji, W Xing, C Hao… [Shanghai Jiao Tong University & Sharpa] (2026) https://t.co/QNGQOak6HJ

X (formerly Twitter)

The Humanoid Hub (@TheHumanoidHub)

강화학습(RL)으로 손 안(in-hand) 스킬(그립 안정성 및 물체 회전)을 학습해 원격조작(teleoperation)을 보조하고, 그 원격조작 데이터를 이용해 힘(force)과 촉각(tactile) 피드백을 융합해 동작을 생성하는 VLA 모델을 학습했다고 보고합니다. 실험은 휴머노이드 하드웨어 'Sharpa'에서 수행되었으며 관련 논문은 arXiv:2603.08122입니다.

https://x.com/TheHumanoidHub/status/2033223493517754828

#robotics #reinforcementlearning #teleoperation #tactile #humanoid

X

X (formerly Twitter)

Ilir Aliu (@IlirAliu_)

SONIC이라는 이름의 오픈소스 프로젝트가 공개되어 누구나 사용할 수 있는 generalist whole-body teleoperation 파이프라인(전신 제어, 기구학 플래너, 원격조작)을 제공합니다. 현재 추론 코드와 모델이 공개되어 있으며, 학습 코드와 지속적 업데이트도 순차적으로 공유될 예정이라는 발표입니다.

https://x.com/IlirAliu_/status/2027669118573244917

#teleoperation #robotics #opensource #wholebody

Ilir Aliu (@IlirAliu_) on X

Generalist whole-body teleoperation for EVERYONE! Now open-source 🔥 Comprehensive pipelines for whole-body control, kinematic planner, and teleoperation: SONIC. And they will all be shared❗️ There will be a continuous update; inference code + model already there, training

X (formerly Twitter)

Here's a photo of one of Icarus Robotics' semi-autonomous teleoperated robots — an important milestone as they build towards fully autonomous systems.

Find out more by listening to Episode 146 of Robot Talk - available on all major podcast providers: https://linktr.ee/robottalkpod

#Robots #Robotics #Space #AI #Drone #ISS #Teleoperation

The Humanoid Hub (@TheHumanoidHub)

휴머노이드 원격조종(teleoperation)은 비용과 확장성 문제가 있는데, 이를 겨냥한 NVIDIA의 EgoScale이 소개됩니다. EgoScale은 수천 시간의 제1인칭(egocentric) 인간 영상으로 사전학습된 VLA 모델이며, 인간-로봇 정렬을 위해 50시간의 인간 데이터와 4시간의 로봇 '플레이' 데이터로 미드트레이닝되었다고 합니다.

https://x.com/TheHumanoidHub/status/2026737482168152258

#nvidia #egoscale #vla #teleoperation #robotics

The Humanoid Hub (@TheHumanoidHub) on X

The problem with humanoid teleoperation is that it is expensive and difficult to scale Enter NVIDIA's EgoScale: - A VLA model pretrained on thousands hours of egocentric human videos. - Mid-trained via 50 hours of human + 4 hours of robot "play" data for human-robot alignment.

X (formerly Twitter)

🎭 Humanoide Roboter wirken autonom sind aber oft ferngesteuert. Teleoperation täuscht, liefert aber wertvolle Trainingsdaten.

👉 Meine Meinung: Kein Trick, sondern sinnvoller Übergang. Problematisch ist nur die fehlende Transparenz.

(Video Credits to Peter Kappes, 28.01.2026, via LinkedIn / LI, "Human Tele-Operators Train Robots with Brain-Control Interface")

#Robotik #KI #Teleoperation

Robert Lufkin MD (@robertlufkinmd)

광저우에 위치한 '완전 로봇화된 조선소'(2025년 예시)를 소개하는 트윗으로, 아직 간헐적 원격조작(teleoperation)이 병행되어 완전 자율은 아님. 어떤 직업들이 자동화될지에 대한 토론을 촉발하는 사례 공유.

https://x.com/robertlufkinmd/status/2012353579969872364

#robotics #automation #china #teleoperation

Robert Lufkin MD (@robertlufkinmd) on X

Daddy, what was a teamster? Fully robotic shipyard in Guangzhou, China 2025. (Note not fully autonomous yet as there is occasional teleoperation) What other common professions are going this way? h/t @lukas_m_ziegler

X (formerly Twitter)

🚫 Wenn Roboter uns zu genau nachahmen...

👉 Meine Meinung: Für Training und Forschung unverzichtbar aber produktiver Einsatz? Noch weit entfernt. Die Entwicklung ist allerdings rasant!

(Video Credits to Wes Morrill, 27.12.2025, via X, "teleoperator kicking himself over not programming Asimov's Laws of Robotics..."; originally seen on the Bilibili account zeonsunlight / BILI; Social Media-Bearbeitung: Endrit Restelica)

#Teleoperation #Robotik #AI

Elon Musk's Optimus Robot Collapses in a Hilariously Suspicious Way

Musk says he's going to make 1 million robots per year by 2030.

Gizmodo
Die neue StVFernLV macht es möglich: Seit dem 1. Dezember dürfen Autos in Deutschland ferngesteuert werden. Kein Fahrer mehr im Auto, sondern Menschen in Leitstellen übernehmen die Kontrolle. #Teleoperation https://winfuture.de/news,155335.html?utm_source=Mastodon&utm_medium=ManualStatus&utm_campaign=SocialMedia
StVFernLV: Ferngelenkte Autos sind jetzt in Deutschland erlaubt

Seit dem 1. Dezember gilt die StVFernLV: Ferngelenkte Autos sind auf deutschen Straßen legal. Doch wie sicher ist die Datenübertragung? Wir schauen auf technische Hürden, die strengen Anforderungen an Telefahrer und die Bedeutung für die Logistik.

WinFuture.de