Unitree@UnitreeRoboticsUnitree G1 has mastered more quirky skills 🤩Unitree G1 has learned the "Anti-Gravity" mode: stability is greatly improved under any action sequence, and even if it falls, it can quickly get back up.
These approaches, Brooks argues, ignore decades of research showing that human dexterity depends on an extraordinarily complex touch-sensing system. He cites work from Roland Johansson's lab at Umeå University showing that when a person's fingertips are anesthetized, a seven-second task of picking up and lighting a match stretches to nearly 30 seconds of fumbling. The human hand contains about 17,000 mechanoreceptors, with 1,000 concentrated in each fingertip alone. Recent research from David Ginty's lab at Harvard has identified 15 families of neurons involved in touch sensing, detecting everything from gentle indentation to vibrations to skin stretching. That's a lot of sensory information that current robot systems cannot yet capture or simulate.
What's that guy in the background holding? At first I thought it was a video game controller because that's how he holds it for most of the video. But it looks more like a tv remote with a lanyard at the end of the video. Emergency off switch maybe?Edit: Or is he providing some sort of general input to the optimus? Advance, arm strike, kick, with the optimus deciding how best to fulfill those inputs?
What's with those weird hands/arms? I assume there are no actuators inside them and they're just mannequin hands/arms?Otherwise, I don't see how they could look like that
Thomas Wolf@Thom_Wolf·Oct 5Wow! Super impressive work by the new Amazon FAR team (from Covariant acquisition).Mapping long sequences of human motion (>30 sec) on robots with a differing shapes or interating with objects (box, table, etc) of different size. Enabling easier in-simulation data-augmentation and zero-shoot transfer.Super impressive and huge help to reduce the need for human teleop data (which is very complex to gather for humanoids) Dataset trajectories on Hugging Face (search OmniRetarget), full code framework to come soonProject page has some pretty three.js interactive demos
Rohan Paul@rohanpaul_ai·Oct 4A Sydney-based startup, Crest Robotics, has built huge spider-like robot, named Charlotte.It's designed to 3D print entire buildings using raw materials it processes on site.And there are ambitions for it to build structures on the Moon.
Quote from: Rohan Paul@rohanpaul_aispider-like robot, named Charlotte.
spider-like robot, named Charlotte.
Optimus busting out the martial arts moves
Nice human-like gait, but how long/far can it run like that?
Quote from: sanman on 12/04/2025 03:11 amNice human-like gait, but how long/far can it run like that?I expect the truly native "robot-optimized" gaits will be all about minimizing wear-and-tear on the mechanical components, and (secondarily) energy optimization.Minimizing total all-cause wear-and-tear is mostly about not falling over (low risk gait), and secondarily about minimizing shock and vibration on actuators and bearings.In a factory context, you don't care if the robots have a "human-like" gait, but you definitely care about your monthly maintenance bill...
...With bipedal walking offering the adaptive dynamic stabilization for irregular terrain, and wheels offering the stability and efficiency of rolling across flat surfaces, then what's the best way to marry them both? We've seen it done with the quadruped-wheeled robots already posted upthread, but what's the best way to do it with a bipedal humanoid form-factor & posture?
Quote from: sanman on 12/04/2025 11:12 pm...With bipedal walking offering the adaptive dynamic stabilization for irregular terrain, and wheels offering the stability and efficiency of rolling across flat surfaces, then what's the best way to marry them both? We've seen it done with the quadruped-wheeled robots already posted upthread, but what's the best way to do it with a bipedal humanoid form-factor & posture?What happens in the startup world quite often is that solutions are created for problems that have not been identified yet - or aren't a problem. The push by Musk et al. for bipedal humanoid robots kind of feels that way sometimes, that the solution is be created before the problem is well defined - at least for Mars.Assuming that bipedal humanoid robots are necessary for colonizing Mars is a great example of this, because we don't yet know what the "jobs" are that need to be solved by robots of any kind, much less bipedal ones.If you need to move material around on the surface of Mars, then a wheeled vehicle may be best. If you need to take rock samples on the side of a steep, rocky slope, then a bipedal humanoid robot may be best. Between those two limited examples I would think that moving material round on the surface of Mars will be a far more predominate need than gathering rock samples from steep slopes.We are still many years away from sending humans to Mars, or even sending advance cargo to Mars. However as we get closer, and the payloads, machinery, and needs become clearer, then we can start figuring out what kind of automated or semi-automated robotic systems we will need.But if you want to debate the solution (i.e. bipedal vs wheeled) then you first need to define the "jobs to be done", so that you have some ability to rank the solutions.
