Northrop Grumman's Mission Extension Vehicle has undocked from an Optus communications satellite after more than a year holding it in position. Four replacement spacecraft launched in July, among them a Mission Robotic Vehicle carrying two robotic arms. The Mission Robotic Vehicle's two arms were developed by DARPA; the other three craft are Mission Extension Pods, smaller modular units.
Researchers at the Karlsruhe Institute of Technology have built a system that predicts the defect in a broken product and disassembles it while protecting the valuable parts. With more than 4 million industrial robots in use worldwide and over 16 million forecast by 2030, the article argues the industry should prepare to dismantle, recycle and rebuild what it makes.
Boston Dynamics says the production Atlas can recognise when its battery is depleted, return to a station, swap the battery in under three minutes and resume its task. It runs about four hours between swaps, or two when lifting heavy loads, while a conventional recharge takes about 90 minutes, which is what makes swapping matter for factories running several shifts.
Humanoid Analytics' July 2026 market snapshot tracks 351 companies and programmes: 225 build complete humanoids and 126 work in adjacent segments such as dexterous manipulation, sensing, components, AI and data infrastructure. That puts 35.9% of the tracked ecosystem into humanoid robotics without building the robot — a count of participants, the firm stresses, not market share. Dexterous manipulation is the largest adjacent segment, with 58 entities.
San Mateo County's Board of Supervisors has unanimously directed staff to draft an ordinance regulating untethered, autonomous or semi-autonomous humanoid robots in commercial settings, while California still has no framework of its own. Staff are to consider requiring that any permitted robot be supervised at all times by a trained human on site, which Humanoids Daily says would undercut remotely teleoperated services such as Tau Robotics' $30-an-hour humanoid cleaning in San Francisco.
World Labs sets out which functional pieces actually compose a world model and what each one is for, arguing the term has become one of the most overloaded in AI. The premise is that language models learn the statistical structure of text while world models must learn the structure of space and time. Computer vision, robotics, reinforcement learning and generative AI each claim to be building world models, and each means something quite different by it.
Flex-π, a 6B-parameter world-action model, is supervised on 3D pointmaps and DINO semantics as well as RGB, all encoded by the same frozen video VAE, and one checkpoint runs on any subset of those streams. The authors report gains of up to 2–7x over the strongest baselines on real bimanual tasks, running faster than π0.5.
DreamX-Phi 1.0 predicts the frames that follow a prescribed sequence of end-effector poses and gripper states, with per-arm geometric encoding so each arm follows its commanded path, and a depth branch and object masks to keep the scene consistent. It took first place on Track 1 and second on Track 2 of the WorldArena 2.0 Challenge; the model and code are to be released.
Researchers in the Czech Republic built microscopic robots from stacked MXene particles that removed microplastics from both soil and water in laboratory tests. Particles under five millimetres are especially hard to recover from soil because they lodge between minerals and organic matter. The MXene particles carry a large surface area and a surface chemistry that attracts plastic, and a nickel coating lets rotating magnetic fields drive them.
Galaxy General Robotics' HumanTracker benchmark scores humanoid motion tracking against 2,500 reference clips retargeted to a 29-DoF Unitree G1-style body, in four families from daily motion to highly dynamic moves. It adds 6,000 human preference pairs comparing rollouts from four trackers — GMT, TWIST2, SONIC and Humanoid-GPT — which train a HumanScore reward model meant to catch foot skating and unstable support that per-frame error misses. It is Apache 2.0, with an ECCV 2026 paper.
Researchers at the RAI Institute and Boston Dynamics describe ZEST, a reinforcement-learning recipe that turns motion capture, single-camera video and keyframe animation into whole-body skills, trained entirely in simulation and run on hardware without fine-tuning. On Atlas, Unitree's G1 and Spot it produced crawls, cartwheels, breakdancing and a backflip, each policy taking about 10 hours on one NVIDIA L4 GPU; so far it is limited to flat, non-slippery ground.
NVIDIA JetPack 7.2.1 adds PyNvVideoCodec 2.2 on Jetson Thor, bringing the hardware video encoder and decoder that JetPack 7.1 opened to C and C++ developers to Python for the first time. NVIDIA calls video a core data path in Jetson applications from robotics to industrial automation, where the engineering lies in choosing the interface, codec, pixel format and memory path, and proving the pipeline meets its latency and throughput goals.
TLabel is a cross-sensor annotation schema for tactile manipulation data, producing 14-dimensional semantic annotations that different sensor types can all emit. Each adapter declares which dimensions it can and cannot annotate, so results from different tactile sensors become comparable rather than sensor-specific. Compliance levels L1 to L4 let each sensor participate at whatever information density it can actually support.
NEURA Robotics is acquiring Bosch Rexroth's ACTIVE Shuttle automated guided vehicle, effective 1 October 2026. Bosch Rexroth's navigation and fleet management will be available through the NEURA platform, with the shuttle gaining NEURA's AI interfaces in stages. The shuttle is already field-proven across industrial customers for moving material between assembly workstations, and NEURA is folding it into a European Physical AI ecosystem.
Agility Robotics plans to deploy Digit V5 to its first customers in December, with safety systems meant to let the humanoid work beside people without the fencing that usually surrounds industrial robots. Its battery is designed for more than 20 hours of work a day, and an ISO-standard flange takes interchangeable grippers, with autonomous tool changes planned. Early deployments are meant to validate the safety claims before wider production.
Global humanoid shipments reached 19,100 units in the first half of 2026, more than three times the 5,100 of a year earlier, and Chinese manufacturers took 97% of them. Agibot overtook Unitree for the top place with 44% of the global total, and the four largest suppliers were all Chinese. Smart Analytics Global puts industrial and commercial work at more than 70% of shipments, and names falling margins and the US import ban as what shapes the next phase.
