Predator is a steel humanoid that got knocked to the floor repeatedly this week & stood back up unassisted every time. Fall recovery is harder than walking, & it's the part worth watching. Upright locomotion is a balance problem with a known starting state. A fall erases that. Predator lands in an arbitrary orientation, has to reconstruct its position from its own sensors, then sequence dozens of joints through a push-up while its center of mass moves the entire time. One error & it's back down. It did this under adversarial conditions, against Frankie LaPenna trying to prevent it. To be clear about what's being measured: REK's founder told NBC News these robots aren't autonomous. They are "100 percent" human pilots in VR headsets. So Predator is testing teleoperation latency, impact tolerance & balance recovery, not machine decision-making. Which makes it a more honest benchmark than most. Lab demos happen on flat floors under controlled conditions. You can't stage a knockdown from someone actively fighting you. Getting up is the gate on humanoids working anywhere near people. Predator cleared it in public, repeatedly, on video. 🦾

A robotic hand that walks away on its own fingers! 🪬 We finally have the Thing from Addams family in robotics, and it was developed at ETH Zürich. No wheels, no legs bolted underneath, no separate drive mechanism. The same fingers that grasp things also carry the hand across a surface and hold up its weight. A quadruped has four identical legs spaced evenly around its body, and most legged-locomotion reward functions quietly assume exactly that. A hand has fingers of different lengths, arranged unevenly around a palm, with a thumb that matches none of them. The team wrote a reward formulation for that unequal geometry, and in simulation it moves faster than tuned rewards borrowed from quadruped work. They kept the hand's existing finger design and its position controller, so this isn't a bespoke machine built for the demo. And power and computation sit onboard, so it runs untethered. → Crawling, steering, and recovery after being knocked over → While holding itself up, it executes a sequence of keyboard commands with no vision at all → It pushes an object to target positions using overhead visual feedback → The simulator was calibrated against measurements taken from the physical hardware A hand that walks to a task and then performs the task is a mobile manipulator with no locomotion hardware in it. Big shoutout to authors @amrhkzp! 😮‍💨 🔗 Here's the paper: ~~  ♻️ Join the weekly robotics newsletter, and never miss any news →

TRON Camp 2026 Documentary EP.04 | When Code Walks Out of Simulation By Day 1, the students had already brought a humanoid robot to life. Across five days, they moved from simulation to real hardware, validating, deploying, and tuning their algorithms on real terrain. What once felt like a long journey from code to reality became a first step, taken by the robot — and by them. TRON Camp 2026 officially comes to a close. See you next year! #LimXDynamics# #TRONCamp2026# #EmbodiedAI# #Locomotion# #Sim2Real#

A dexterous hand that stands up and does pull-ups! 😮‍💨 This is Handroid that is a gripper that can transform into a little humanoid robot. It was developed across teams from The University of North Carolina at Chapel Hill and Stanford University. It looks like The Thing and I'm totally obsessed with it! 🫳🏼 It's a desktop-scale robot with 27 degrees of freedom that runs as a dexterous hand when manipulation is needed, and as a humanoid when locomotion is the prio. The secret sauce is in the morphology. The finger joints double as humanoid limbs, and the same compact actuator modules get reassigned across fingers, arms, legs and base actuation. P.S. Imagine your grippers being able to come down and start pushing things :D 🔗 Here's the project page: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

NEW SKILL: This team take humanoid parkour to the next level! This project involves @zhenkirito123, @Yuanhang__Zhang, @pabbeel, @carlo_sferrazza, @GuanyaShi, and others. Called Perceptive Humanoid Parkour (PHP), it lets a Unitree G1 humanoid (29 DoF, 1.3 m) run long-horizon, vision-based parkour and decide on its own whether to step over, climb onto, vault, or roll off obstacles, using only onboard depth sensing (a 30 Hz camera) plus a discrete 2D velocity command from the operator. It builds the motions by motion matching, a nearest-neighbor search in a 27-dimensional feature space (imported from the video-game industry) that stitches retargeted atomic human clips into long-horizon kinematic trajectories through a shared Locomotion to Skill to Locomotion manifold. It then trains per-motion RL tracking experts (with a privileged height scan) and distills them into one depth-based multi-skill student via DAgger combined with RL. It replaces hand-scripted skill sequencing and operator-triggered skill selection. The onboard depth-only student holds 0.95 to 1.00 success across obstacle heights in simulation, while a naive end-to-end depth policy collapses from 0.95 to 0.07/0.08 at 58 and 76 cm. The upside comes from distilling privileged-height-scan experts and from motion-matching composition, not from training onboard from scratch. The operator sends only a coarse 2D velocity (W/A/D keys), and the robot itself picks which skill to run from the depth image. The training data contains only single-obstacle traversals, so the ability to chain skills across a multi-obstacle course is generalization the policy was never explicitly taught. Transitions are constructed in kinematics, not rediscovered by learning. Every skill enters and exits through a shared locomotion manifold, so no hand-captured skill-to-skill transition clip is needed. Motion matching densifies a sparse library into smooth long-horizon trajectories before any RL. The entire parkour library is only about 66 seconds of mocap, and each skill is just a few seconds. Ablating approach-distance variety ("Extreme Distances") drops the 76 and 94 cm climbs to 0.62 and 0.64, and "Half Density" drops the 76 cm climb to 0.32. Motion matching's job is to manufacture stride-phase and approach variety out of seconds of data. It clears a 1.25 m wall (96 percent of its 1.3 m height) in 3.63 s, and a cat-vault peaks at 3.41 m/s clearing a 0.4 m by 0.5 m obstacle in 0.8 s. A single 30 Hz depth stream drives both traversal and runtime re-planning, shown by displacing obstacles about 0.5 m mid-run.

NEW RESEARCH: You can play soccer with your Unitree G1 Humanoid! This project involves @PeterYChong, @YuhangLu2000, @sephy_li, and others. It is a collaboration between OpenDriveLab, the University of Hong Kong, the Chinese University of Hong Kong and Archon Robotics. Called RoboNaldo, it is a reinforcement-learning policy that makes a Unitree G1 humanoid (29 DoF, 35 kg) shoot a soccer ball, both stationary free-kicks and one-touch shots on a moving ball. It is trained by a three-stage motion-guided curriculum in Isaac Lab (4096 parallel environments): - Stage 1 imitates a single human-kick reference (retargeted from a human video via GVHMR and GMR) with no ball or task reward - Stage 2 adds the ball, a target and shooting rewards while randomizing the ball spawn - Stage 3 adds a high-level locomotion command and a kick trigger so it can walk up and volley Perception is onboard and egocentric: a head-mounted Livox MID-360 LiDAR and a chest RealSense D435 track a retro-reflective ball (Kalman-fused), and an AprilTag (or a fixed coordinate) marks the target, all at 50 Hz with no motion capture. What is interesting imho: if you remove Stage 1 (imitating one retargeted human-kick video), the alive rate collapses to 1.2 percent, it topples on nearly every attempt. The human prior's real job is whole-body stability through a violent, high-impulse motion, not style; the hard part of a robot kick is not aiming but not tipping over from your own swing The onboard, no motion capture perception still needs a retro-reflective ball and an AprilTag goal. Ball localization is a head Livox MID-360 LiDAR plus a chest RealSense IR camera tracking a retro-reflective ball, fused by a Kalman filter. The target is read from an AprilTag or hardcoded as a fixed coordinate. It shoots hard but sub-human and only roughly accurate. Peak ball velocity is 13.10 m/s, which it reports as 59 to 71 percent of a professional open-play shot (about two-thirds of a real Ronaldo, fitting the pun). Stationary accuracy is 0.73 m average error from 3 m, with 31.5 percent of shots inside 0.5 m and 80.6 percent inside 1.0 m. The entire motion prior is one human video. There is no kick dataset and no mocap suit: a single human-kick clip retargeted to the 29-DoF G1 via GVHMR and GMR, then curriculum RL across 4096 Isaac Lab environments deviates from it for aim, contact and timing. It is the same one-demonstration-plus-sim-compute recipe as LadderMan and the perceptive-parkour projects I talked about previously. Also fun fact: it is an autonomous-driving lab pivoting to humanoid soccer. RoboNaldo comes from OpenDriveLab (the HKU end-to-end self-driving / UniAD group) with CUHK and the startup Archon Robotics, all Hong Kong based.

29 builders. 13 countries. One mission. We're excited to announce the 29 developers selected for the first cohort of Project EDGE. Over the past few weeks, we've received applications from researchers, robotics engineers, open-source contributors, startups, creators, and universities around the world. After reviewing each application, we've selected our first group of builders to join the Project EDGE Founding Cohort. This cohort includes: • Leading university robotics labs • AI researchers and robotics engineers • Open-source robotics contributors • Robotics startups and infrastructure builders • Developer communities and robotics creators Together, they'll explore humanoid locomotion, robot learning, teleoperation, ROS 2, embodied AI, data pipelines, and real-world deployment on LUMOS NIX. This is just the beginning.Project EDGE is still accepting applications, and we'll continue selecting outstanding builders from around the world. If you're building the future of embodied AI, we'd love to hear from you. Apply here:

Will Embodied AI Be AI’s Next Breakout? At this week’s World Humanoid Robot Games, organizers released a dataset containing more than 2,500 hours of real-world robot operation data. It reflects a broader shift: embodied AI is moving beyond stage demos and toward data accumulated through physical interaction. Zhihu contributor 任广杰, Assistant Vice President at @LejuRobotics_ , believes embodied AI will become AI’s next breakout field. But it will not resemble ChatGPT’s overnight rise. The real signal will be a steady increase in the hours robots can work reliably in real environments. 1️⃣ Physical AI plays by different rules The previous AI waves, including computer vision and NLP, operated largely in information space. When a model makes a mistake, the task can usually be retried. Robots operate in the physical world. Changes in lighting, friction, payload, or object position can damage equipment or halt a production line. That creates two requirements AI has rarely had to meet at scale: 🔹 Real-time, closed-loop control 🔹 Consistently high success rates in uncontrolled environments A successful demo is not enough. The system must continue working when conditions change. 2️⃣ Three inflection points have arrived 🔹 Robot bodies are becoming reliable enough Integrated actuators now combine motors, reducers, encoders, and drivers. This has sharply reduced joint failure rates. At Leju, robot mean time between failures has increased from hours to thousands of hours. Below that threshold, the economics do not work. Maintenance and downtime can erase the value of replacing human labor. 🔹 Robots now have both a “brain” and a “cerebellum” Traditional industrial robots execute fixed instructions. A change in task usually requires new programming and teaching. Large models give robots task-level understanding. A robot can inspect randomly placed components, then plan the order and pose for grasping them. Meanwhile, lower-level control systems handle balance, locomotion, and physical disturbances. The “brain” decides what to do. The “cerebellum” keeps the movement stable. The body must survive repeated execution. 🔹 The data flywheel is starting to turn Embodied AI is beginning to follow the path of autonomous driving. Real robots generate operational data. That data improves the models, which then return to the physical world and generate more useful experience. This closed loop is essential because many problems only appear during deployment. 3️⃣ The real gap lies between the lab and the factory Leju began developing humanoid robots in 2016. One of the clearest changes in the current cycle is that factories are now willing to open real production stations for testing. Three years ago, that was much harder. A 95% grasping success rate may look impressive in a laboratory. A factory may require 99% availability under changing lighting, reflective surfaces, deformed containers, and shifting object positions. Customers therefore ask very practical questions: 🔹 How long can the robot operate without failure? 🔹 Can it recover from an exception by itself? 🔹 Can it keep pace with the production line? Answering them requires the body, controller, model, data loop, and deployment engineering to work as one system. 4️⃣ The first commercial wave is already taking shape The earliest deployments appear in structured environments with clear task boundaries and measurable value. Leju’s full-size Kuavo 5 robots are already used for reception and guided tours in exhibition halls, banks, and stores, as well as for park inspection. The company says these projects now cover 23 Chinese provinces and hundreds of customer cases. In manufacturing, Leju explored more than 70 factories and hundreds of potential tasks over the past year. Dozens passed proof-of-concept acceptance. Its wheeled Kuavo 5-W has entered small-batch deployment for tasks such as loading small automotive components and depalletizing cartons. The important change is that some customers are beginning to pay for repeatable workflows, not merely demonstration projects. 5️⃣ Home robotics remains the endgame Homes may eventually become the largest market, but they are also the hardest environment. They are highly unstructured. Task boundaries are almost unlimited, and consumers are extremely sensitive to cost. Two conditions must be met: ✅ Embodied models must generalize well enough for real household use. ✅ The total cost of the robot must fall below the cost of human labor. Until both happen, industrial and commercial settings will remain the more realistic path to scale. 6️⃣ Watch effective working hours, not flashy demos The best measure of an embodied AI system is how long it can create value without human intervention. Three questions matter: 🔹 Can it do the job? This depends on whether the task can be decomposed and whether the environment is sufficiently structured. 🔹 Can it do the job reliably? The key metrics are continuous operating time and autonomous recovery from failures. 🔹 Is it economically worthwhile? The customer must be able to calculate the full lifecycle cost. Once the economics become clear, repeat purchases can begin. Embodied AI is already moving from exhibition piece to production tool. Its breakout will not be defined by one viral robot. It will be defined by thousands of machines quietly accumulating useful working hours in factories, stores, and eventually homes. 🔗 Full analysis: #EmbodiedAI# #HumanoidRobots# #Robotics# #PhysicalAI# #ArtificialIntelligence# #ChinaTech#

NEW STARTUP: Retrofit your construction equipment to make it autonomous! Bedrock Robotics @BedrockRobotics is a 2024 San Francisco startup, founded by former Waymo engineers (CEO Boris Sofman, ex-head of Waymo's trucking program), that retrofits autonomy onto existing construction machines rather than building new ones. Its product, the Bedrock Operator, is a kit of lidar, GPS, motion sensors, cameras and onboard computers with AI software that mounts to a machine in a few hours with no permanent modification, letting excavators, bulldozers, loaders, graders and compactors navigate and do earthmoving work on jobsites without a human in the cab. It is deployed with US general contractors (a named customer is Champion Site Prep in central Texas), and in 2026 it raised a $270M Series B co-led by CapitalG and the Valor Atreides AI Fund at a $1.75B valuation (over $350M total), targeting its first fully operator-out excavator deployments later in the year. I find it interesting that the retrofit approach introduced by self-driving found its product-market fit in construction. Bedrock was founded in 2024 by ex-Waymo engineers (Sofman ran Waymo's trucking program, and recent hires include a head of people from Waymo and a head of evaluation from Meta AI safety), and it points the same perception-and-planning stack at construction. Sites are private, geofenced, low-speed and pedestrian-free, with a far lower safety and regulatory bar than public roads and an immediate customer ROI. The goal here is the operator-less excavator, because a dozer drives but an excavator manipulates. Graders, compactors and dozers mostly move in patterns, close to the lane-following the AV founders already mastered. An excavator digs, swings and dumps with a multi-DOF hydraulic arm on unstructured soil, which is articulated manipulation, not locomotion. Bedrock explicitly calls excavators "complex, articulated machines" and targets its first fully operator-out excavator deployment in 2026, the hard frontier of the category. The Bedrock Operator mounts in a few hours with no permanent modification to a machine the contractor already owns, so there is no warranty risk, no fleet replacement, and it can be removed. Heavy equipment lasts fifteen to twenty years, so building autonomy into new machines leaves the existing fleet untouched for a decade; a removable kit captures that installed base today, the same moat Gravis is chasing on the other continent. A $1.75B valuation on a company targeting its first operator-less deployment is AI-frontier pricing, and the backers are AI and growth money (CapitalG / Google, NVIDIA, 8VC, Eclipse, a Valor "AI Fund") plus, tellingly, Tishman Speyer, a giant real-estate developer, and MIT. The developer money is customer-adjacent, a signal of pull from the people who actually build, while the named live customer (Champion Site Prep, central Texas) grounds it in real jobsite deployment rather than a demo.