Lukas Ziegler (@lukas_m_ziegler)
콜롬비아대 연구진이 LeRobot의 SO10X 플랫폼에 FlexiTac 촉각 센서를 추가했다. 적은 양의 촉각 정보만으로도 ACT, Diffusion 등 여러 정책 계열에서 접촉이 많은 조작 작업의 성공률이 유의미하게 향상됨을 확인했다.
Adding tactile sensing to LeRobot! Columbia University researchers added FlexiTac tactile sensing to the SO10X platform. They found that even a modest amount of touch meaningfully lifts success rates on contact-rich manipulation across four policy families: ACT, Diffusion
Lukas Ziegler (@lukas_m_ziegler)
연구진이 5달러 수준의 저비용 오픈소스 자기식 촉각 센서를 공개했다. 제작이 쉽고 촉각 감지를 민주화할 수 있어, 가정·사무실 같은 비정형 환경에서 작동하는 로봇의 힘/접촉 감지 능력을 크게 향상시킬 수 있는 주목할 만한 로보틱스 기술이다.
Open-source magnetic tactile sensor for $5! 🧲 Researchers introduced a magnetic tactile sensor that's low-cost, and easy to fabricate, democratizing tactile sensing for robotics. Operating in unstructured environments like homes and offices requires robots to sense forces
RE: https://mastodon.social/@stemsearchgroup/116370623815342121
So much wrong with this and I haven't even clicked through yet.
A) "Feel" is a function of sensors, not AI, and people have been working on tactile sensors with a wide range of properties for years. The TacTip that senses direction of force as well as pressure. The artificial skin. The Festo tentacle gripper. The jamming gripper (coffee grounds in a balloon is makes a surprisingly good gripper!).
B) A robot doesn't "go blind" when it touches something. All its other sensors still work, plus whatever sensors it has specifically for contact and touch (and yes, it needs those).
C) There has been tons of work in this area that doesn't involve "AI" at all. Traditionally, this is both a hardware problem and a software problem. And Gill Pratt said "grasping is solved" at IROS in Portugal in 2012.
D) WTF is "physical AI"? We have had "autonomous robots" "intelligent agents" "cyber-physical systems" - now we have yet another stupid term foisted on us by the "robotics is a subset of CS" crowd? Again? If it's AI, it's software. If it has a body and it's trying to pick things up without destroying them, it's a robot.
Lukas Ziegler (@lukas_m_ziegler)
SharpaRobotics가 촉각 센서를 탑재한 로봇 손 Wave를 공개했다. 인간 손을 닮은 isomorphic 디자인과 22개의 능동 자유도를 갖춰, 강한 파지력과 빠른 동작, 정교한 조작을 동시에 구현하며 인간 수준의 손재주를 목표로 한다.
Dexterous hand with tactile sensing! 🤲 @SharpaRobotics recently unveiled Wave, a dexterous robotic hand that combines strength, speed, and advanced tactile sensing. The isomorphic design mirrors the human hand with 22 active degrees of freedom, unlocking human-level dexterity
The Humanoid Hub (@TheHumanoidHub)
UC Berkeley 교수이자 Amazon의 Distinguished Scientist인 Jitendra Malik가 병렬턱 그리퍼(parallel-jaw grippers)가 정교한 조작에 적합하지 않다고 비판하며, 촉각 센싱을 갖춘 다지(多指) 핸드가 고급 수준의 외과적·정교한 조작을 위해 필요하다고 주장하는 내용입니다.
https://x.com/TheHumanoidHub/status/2023138332952363354
#roboticmanipulation #dexterity #tactilesensing #jitendramalik
Jitendra Malik rants about parallel-jaw grippers being inadequate; he believes multi-fingered hands with tactile sensing are necessary for advanced dexterous manipulation. Malik is a Professor at UC Berkeley and a Distinguished Scientist at Amazon.
Amirhosein Shirani (@amirhshirani)
MSFT Research의 Rho-alpha와 VLA+로의 진화가 촉망된다는 내용으로, 특히 촉각 피드백(tactile feedback) 추가가 주목받고 있음. 다만 저자는 데이터 수집뿐 아니라 촉각 센싱의 정확도와 지연(latency)이 정밀한 미세 동작 작업에서 주요 공학적 난제라고 지적함.
The evolution to VLA+ with @MSFTResearch’s Rho-alpha is promising, particularly the addition of tactile feedback. Yet, the primary engineering hurdle isn't just collecting data—it is the accuracy of that tactile sensing. For tasks requiring fine motor skills, even minor latency
Ronald van Loon (@Ronald_vanLoon)
ORCA v1이라는 이름의 신형 휴머노이드 핸드가 소개되었다. 17자유도(17-DoF) 텐던 구동 구조(tendon-driven), 촉각(tactile) 센싱, 팝가능한(poppable) 조인트를 특징으로 하며, 로보틱스와 감각 피드백을 강조한 하드웨어 신제품(연구·개발) 발표로 보인다.
Amazon Debuts Vulcan Robot with Tactile Sensing for Warehouse Automation
#Amazon #Robotics #Vulcan #Sequoia #AI #Automation #WarehouseAutomation #FulfillmentCenter #FutureOfWork #TactileSensing #LogisticsTech #Robots #Innovation #AmazonRobotics
Day two of the Touch Sensing and Processing Summer School is complete. Today the participants had the opportunity to go on an excursion to Saxon Switzerland, after which they could see Dresden's old town from a historic steamboat and were given a short tour around the city. The weather was on our side. Tomorrow the programme continues with a series of talks.
Kick-off for the Touch Sensing and Processing Summer School! Roberto Calandra welcomes the 40 researchers who have travelled here from 14 countries. Today, the focus is on the significance of touch sensing and processing as well as the most important challenges. Furthermore, Zaki Hussein gives us insights into his journey and the industry through his company Touchlab Limited. Will Touch Sensing be the next breakthrough in robotics?
sayzard
Robotistry
sayzard
Winbuzzer
SECAI