Sortation robots at adidas warehouse! 👟 DHL Supply Chain has deployed sortation robots from Unbox Robotics at its B2C warehouse in India. These robots are designed to automate marketplace and transporter sorting processes and auto-manifest generation, replacing labor-intensive manual tasks. 📦 Each robot is configured to handle 31 unique... sort destinations and 48 bin positions, delivering over a 25% improvement in operational efficiency. Integrated with Warehouse Management and Carrier Management Systems, the robots ensure precise sorting with high accuracy, reducing delays and enhancing productivity. Hypnotizing to watch! That's why I love robotics so much! 💛show more

Robot 🦾🦾

31 sort destinations, zero job security. Canada’s still debating basic policies while automation steamrolls local labor markets. Grim. 🤖

If you love them so much, why don't you marry them, Lukas?

looks like there is still a lot imporvement potential.

Robotics in action—looks like a busy, productive spot!

Unitee G1 has taken over people's living area 😂It chases and plays with the little girl. But she doesn't seem scared.

Quadrupeds are fast. Agile. Great at locomotion. But can they manipulate? A new approach from Carnegie Mellon University, Google DeepMind, and Bosch is teaching quadrupedal robots to do more than walk, they’re learning to interact. It’s called Human2LocoMan: a system that uses human data to pretrain robot policies before finetuning on real hardware. The result? A four-legged robot that can walk, carry, organize, scoop, and sort with both single and dual-arm control. By pretraining on human motion, they cut the amount of robot data in half—while improving success rates by over 80% in unfamiliar environments. Their Modularized Cross-Embodiment Transformer (MXT) learns from both human and robot demonstrations, then generalizes those skills to physical tasks—no hardcoded behaviors required. It’s locomotion and manipulation. A quadruped that can walk and clean up after itself?

Wow, incredible to pass $1M in orders since launch on July 1st! We launched K-Bot to give the world an open-source humanoid robot anyone can own. Thank you to everyone who ordered! The Humanoids revolution just got started 🚀🚀🚀

I can’t believe it worked! The self driving screwdriver is alive! This was the first successful autonomous inference, fifth attempt total, and first model! I trained ACT for 10 hours on 100 episodes of training data I collected over the weekend.. I did not clean the dataset, I know there are about 10-15 bad episodes in the dataset so the fact that the first model trained on this data done this well was somewhat surprising. My train val losses looked good and did not indicate overfitting (I ran validation on 25 of the episodes). I don’t have a sim for this robot so I wasn’t able to eyeball the robot in sim during training. Lots of improvements to make: clean data, gather more data, data augmentation, add better metrics… but this is looking very promising so far. I have left the video unedited, for science @huggingface @LeRobotHF is the software used here for hardware integration, data collection, model training and inference. The arm is the Alex Koch low cost robotic arm v1.1 mod by Jess Moss, and further modified with the screwdriver tool by yours truly.