Robotics

Robotics

@robotics.bsky.social

Robotics research.

1.9K followers167 following42 posts

Robotics reposted

No GPS, no light, no map. Philadelphia's Exyn just joined a national lab team under the DOE's $29.5M Mine of the Future push, sending autonomous drones to find critical minerals underground. Its drones already cut survey scans from 40 minutes to 5. Source: Exyn. Video: Exyn Technologies #Robotics

Robotics reposted

Factory robots that retool in minutes, not weeks. Industrial Next (SF) builds AI work cells that learn new tasks from simulation and demo instead of months of custom programming. Now BMW is evaluating them for car production. Source: TechCrunch. Image: Industrial Next #Robotics #ManufacturingAI

Robotics reposted

OpenAI just joined the humanoid race for real. Sam Altman confirmed on a podcast: "We will" build a humanoid, with a demo by 2027. His thesis: the brain matters more than the body. Hiring is underway across actuators, motion control and supply chain. Source: Engadget #Robotics #Humanoid #PhysicalAI

Robotics reposted

$400M says the runway is optional. The Navy put $150M more into Shield AI's X-BAT, the AI-piloted fighter jet that takes off vertically from a ship's deck. Shield matched it with $150M of its own. Flight tests start this year. Source: Shield AI. Video: Shield AI #Robotics #DefenseTech

Robotics reposted

RP1 in motion. RoboParty's fully open-source humanoid takes kicks, recovers its balance, and runs on their UFO reinforcement learning stack. Open hardware is the headline, but open control is the real story. Source: RoboParty. Video: RoboParty #Robotics #OpenSource #Humanoid

Robotics reposted

Open-source humanoids just got real. RoboParty launched RP1: open hardware, open actuators, open SDK. When the robot body is a platform anyone can build on, innovation stops belonging to five labs. Source: RoboParty. Image: RoboParty #Robotics #OpenSource #Humanoid

Robotics reposted

Robotics is rewriting the US economy. McKinsey: AI agents and robots could add $2.9T a year by 2030. IFR: US robot installs up 11% in 2025, but China still installs 10x more than the US. The numbers behind the build-out. Source: IFR, McKinsey. Video: ROBOTICS.biz #Robotics #USEconomy

The Austin Robotics & AI Meetup is looking for speakers! If you have a robotics topic you'd like to share, we want to hear from you. The meetups are online, making it easy to present from anywhere. Ready to give a talk? Learn more here: www.meetup.com/austin-robot...

meetup.comAustin Robotics & AI | MeetupRobotics & AI for Robotics tech community here it the heart of Texas. Join our Austin community to learn and connect with other local tech people working in the field for decades or just getting start...

Robotics reposted

Robotics reposted

Instant Policy from Vitalis Vosylius and Edward Johns​ of ICL enables *one shot, in-context learning* of new robotics tasks. Thats the dream. (Just won a well deserved best paper at ICLR robot learning workshop) ​

Robotics reposted

Robotics reposted

Robotics reposted

ovo.bot LeRobot SO100 robot arm now shipping from the USA! 🇺🇸🇺🇸🇺🇸 If you're in the US, order your DIY kit or fully assembled robot and start experimenting with the latest robotics advancements with minimum hassle. #robotics #lerobot #robot #huggingface #ai #robot #robots #opensource

Robotics reposted

big congrats to Huggingface @hf.co acquiring Pollen Robotics! HF has done a great job stewarding open-source robotics and now will have humanoid hardware to go with it!

DIAL-MPC: A training-free MPC framework enabling real-time torque-level control over full-order legged robot dynamics. Achieves 13.4× lower tracking error than standard MPPI and outperforms RL by 40% in challenging jumps—no training required! lecar-lab.github.io/dial-mpc/ #Robotics #AI #RL

lecar-lab.github.io Full-Order Sampling-Based MPC for Torque-Level Locomotion Control via Diffusion-Style Annealing Full-Order Sampling-Based MPC for Torque-Level Locomotion Control via Diffusion-Style Annealing

Paper introduces a memory-based learning approach that enables robots to tackle complex, constraint-rich tasks more efficiently. Using symbolic reasoning and temporal action ordering, method outperforms RL techniques by an order of magnitude. 📄 Read more: arxiv.org/abs/2303.04327 #Robotics #AI

arxiv.orgUsing Memory-Based Learning to Solve Tasks with State-Action ConstraintsTasks where the set of possible actions depend discontinuously on the state pose a significant challenge for current reinforcement learning algorithms. For example, a locked door must be first unlocke...

🤖 New from Stanford, UC Berkeley & Meta: robots mastering fine bimanual tasks—like threading zip ties & slotting batteries—with low-cost hardware and just 10 minutes of demos.  Introducing ACT: Action Chunking with Transformers.  📄 Read more: arxiv.org/abs/2304.13705

arxiv.orgLearning Fine-Grained Bimanual Manipulation with Low-Cost HardwareFine manipulation tasks, such as threading cable ties or slotting a battery, are notoriously difficult for robots because they require precision, careful coordination of contact forces, and closed-loo...