Robotics Weekly Review 2026-06-28
Week In Review
This was a week defined by industrial robotics’ transition from spectacle to infrastructure. NVIDIA used the opening of Automate 2026 in Chicago to release Halos for Robotics, a full-stack functional-safety system intended to do for fenceless humanoids what its automotive Halos work did for self-driving cars. Within hours, a roster of European robot makers — Wandelbots, Neura, Extend Robotics, Schaeffler — said they were building on the same Isaac, Omniverse and Halos stack. The week’s other major show-floor news, ABB’s Physical AI Toolchain and Kawasaki’s expanded collaboration with Dexterity, pointed in the same direction: shared safety primitives, shared simulation platforms, and shared world-model software underneath whatever brand of arm or torso a factory eventually buys.
Financial markets ratified the shift the same week. Agility Robotics announced a $2.5 billion SPAC merger with Churchill Capital XI, becoming the first US-listed pure-play humanoid company and the first deployment partner for NVIDIA Halos for Robotics. The booth-level counterpart to that capital story was Mantis Robotics’ new dual-arm fenceless system, aimed squarely at the same retail and light-industrial deployments where humanoids are starting to appear — a reminder that “Physical AI” in 2026 covers a wider design space than the bipedal robots that get most of the press.
Medical robotics had its own coordinating week in London, where the Hamlyn Symposium on Medical Robotics opened with a deliberate focus on getting validated systems into low- and middle-income healthcare, and the Surgical Robotics Industry Awards 2026 recognized winners that ranged from India’s SS Innovations to Microbot Medical and Moon Surgical — a noticeably more globally distributed slate than the field had a few years ago. Research news rounded out the week: MIT presented a low-power chip that lets centimeter-scale UAVs build their own 3D obstacle maps on roughly the budget of a single LED — the kind of foundational compute work that determines what autonomous systems can plausibly be built at the small end of the size spectrum. Taken together, the week showed an industry quietly settling into the unglamorous middle of an S-curve: standardizing the layers below the product, and starting to be priced like infrastructure rather than science fiction.
Items
Agility Robotics to Go Public via $2.5B SPAC Merger
Agility Robotics, maker of the Digit warehouse humanoid, announced on June 24 that it will combine with Churchill Capital Corp XI to list on Nasdaq at an enterprise value of about $2.5 billion. The deal makes Agility the first pure-play humanoid robotics company on US public markets and gives the sector a daily, quoted price for the first time.
Agility’s financial disclosures lean on its track record of paid deployments rather than promised demos. Digit robots are already working in Amazon and GXO warehouses; by late 2025, GeekWire’s review of the filings notes that Digit had moved more than 100,000 totes at GXO’s Flowery Branch, Georgia facility, the most documented humanoid commercial deployment of any company. Agility says it has more than $300 million of multi-year orders booked for the next-generation Digit v5, with a pipeline of over 30 customers.
The listing is also a vote on the architectural turn the rest of the week underlines. Digit v5 is being designed for “cooperative safety” — working in shared space with people without a fence or workcell — and Agility is the first announced integration partner for NVIDIA Halos for Robotics. A public market valuation for an unfenced, AI-controlled humanoid will become a recurring data point for everyone else trying to scale.
Source: Agility Robotics
NVIDIA Announces Halos for Robotics, a Full-Stack Physical AI Safety System
On June 22, NVIDIA introduced Halos for Robotics, extending its automotive Halos safety framework to robots and other physical-AI systems. The company describes it as the industry’s first full-stack, standardized safety architecture covering chips, system software, simulation, validation and reporting in one stack — the layers that have historically been bolted together separately by each robotics vendor.
NVIDIA frames Halos for Robotics as the harvest of more than 18,000 engineering-years of work on autonomous-vehicle safety, retargeted at humanoids and industrial robots that increasingly share floor space with people. The stack includes safety-rated compute, runtime monitors, an inspection lab that issues compliance reports against standards like ISO 13482 for personal-care robots, and Omniverse-based simulation for pre-deployment validation.
The strategic significance is less about any single tool and more about who controls the safety substrate as humanoids scale. If most robot makers adopt the same Halos primitives — as Agility is doing first, with Schaeffler, Wandelbots and others queuing up — the industry effectively converges on a shared safety vocabulary in the same way autonomous-vehicle work converged on ISO 26262. That makes regulator engagement, insurance underwriting and customer due-diligence dramatically easier than the bespoke patchwork of the last decade.
Source: NVIDIA Newsroom
European Robot Makers Adopt NVIDIA’s Isaac, Omniverse and Halos Stack
Alongside the Halos announcement, NVIDIA detailed a slate of European industrial-robotics adopters lining up on the same software stack. Wandelbots is integrating its NOVA Operating System with Omniverse so that customers can simulate, validate and tune robot behaviors virtually before pushing them to physical hardware. Extend Robotics is building on Isaac GR00T to let operators teleoperate and train robots for niche jobs like nuclear-decommissioning manipulation. Neura Robotics is post-training the open Isaac GR00T N1 foundation model with NVIDIA’s GR00T-Mimic toolchain to teach its service robot MiPA new skills from a small number of human demonstrations.
The names matter. Schaeffler, an automotive-tier supplier with a global manufacturing footprint, is starting its Omniverse rollout for production-line digital twins; Bosch is plugged into the Halos AI Systems Inspection Lab to extend its driver-assistance sensor expertise into broader physical-AI safety validation. These are exactly the European industrial heavyweights whose adoption choices set the de-facto standards their suppliers eventually have to match.
The pattern is reminiscent of CUDA in scientific computing or Kubernetes in cloud infrastructure: a vendor stack moves from interesting option to gravity well as the most reputable users sign on. For European robotics — which has historically been fragmented across national champions and university spinouts — this is the first time a single horizontal software stack appears poised to anchor the region’s robot fleet development.
Source: NVIDIA Blog
ABB Robotics Unveils Industry Physical AI Toolchain at Automate 2026
ABB used Automate 2026 to publicly show its complete Physical AI Toolchain for the first time — a software stack that spans synthetic data generation, model training, validation, deployment and ongoing optimization for robot AI. The toolchain is the production-side counterpart to RobotStudio HyperReality, the simulation environment ABB and NVIDIA announced in March that integrates Omniverse libraries and runs ABB’s virtual controller on the same firmware as its physical robots, with what ABB cites as roughly 99% sim-to-real transfer accuracy.
ABB also debuted hardware to go with the software, including the new PoWa cobot family — a higher-payload, higher-speed collaborative arm line — and an AI-powered palletizing cell that handles randomized inputs without pre-programmed instructions, the kind of capability that has long limited cobot palletizing to highly structured loads.
Stitched together, ABB’s announcements illustrate the same pattern as NVIDIA’s: the industry is unbundling robot AI into reusable layers. A simulator that matches the real controller, a training stack that feeds it, a fleet of cobots that use the resulting policies, and an integrator ecosystem that ties them into existing warehouses. ABB is positioning itself to sell into every layer of that stack rather than only the arm.
Source: Robotics Tomorrow
Kawasaki Robotics and Dexterity Expand Mech Humanoid Warehouse Collaboration
On June 23, Kawasaki Robotics and Dexterity announced an expanded partnership around Dexterity’s Mech “super-humanoid” warehouse robot, built on Kawasaki’s new RL030N 8-DoF industrial arm. Dexterity is scaling production of Mech specifically for trailer loading and unloading — two of the highest-cost, hardest-to-staff jobs in third-party logistics — and Kawasaki is the industrial-arm partner producing the underlying hardware at volume.
Mech runs Dexterity’s full Physical AI software stack, including a Foresight World Model that lets the robot anticipate how parcels and pallets will move before it touches them. The system is designed for the unstructured, variable-mass items typical of real trailers, where conventional pick-and-place automation has struggled. Pairing that with Kawasaki’s industrial-grade arm — designed for high uptime and serviceability — is meant to push the system from pilots into multi-site fleet deployments.
The partnership is a useful data point on what “humanoid robot” actually means at scale. Mech is not a bipedal walker; it is a torso-and-arms platform optimized for the specific physical envelope of a parked trailer. Many of the warehouse jobs that look in news headlines like they require Atlas-style locomotion are in fact dual-arm manipulation problems standing in one place — and the leaders in that segment are quietly converging on that form factor.
Source: PR Newswire
Mantis Robotics Launches Dual-Arm Fenceless Robot for Industrial and Retail Automation
On June 22, Mantis Robotics announced a new dual-arm robot designed from the start to operate fenceless alongside people — targeted not only at industrial cells but at retail and back-of-house environments where caging a robot is impractical. The system combines two Mantis arms with the company’s existing sensor and safety architecture, and is being positioned for tasks ranging from bin picking to point-of-sale restocking.
The technical bet is that Physical AI advances now make it possible to demonstrate, rather than just argue for, the safety case for unfenced multi-arm operation. Mantis’s launch coincides with NVIDIA’s Halos for Robotics rollout, and the broader industry move to standardize the safety primitives that let regulators and insurers sign off on robots working in human-occupied spaces.
The retail angle is the more novel part. Industrial fenceless cobots have existed for a decade, but applying the same model to a stockroom or back of a store opens up an automation customer base that has historically been priced out of robotics integration. If dual-arm fenceless systems can prove themselves there, the addressable market expands well beyond the manufacturing customers most robotics companies have focused on.
Source: PR Newswire
Automate 2026 Opens in Chicago with NVIDIA-Sponsored Humanoid Pavilion
Automate 2026, the Association for Advancing Automation’s flagship show, opened June 22 at McCormick Place in Chicago and ran through June 25. Organizers reported more than 50,000 attendees, over 1,000 exhibitors and 450,000 square feet of floor — making it the largest edition in the show’s 50-year history.
The most-watched addition this year was a dedicated Humanoid Robot Pavilion, sponsored by NVIDIA, featuring more than 20 humanoid robots and related organizations from around the world. The accompanying Humanoid Robot Forum on June 23-24 brought together engineers, integrators and policy people to talk about deployment and commercialization rather than the demonstration videos that dominated the discussion two years ago.
Beyond the pavilion, the keynote agenda — featuring leadership from Cognex, FANUC, Intrinsic, Schneider Electric, Siemens Digital Industries and Standard Bots — reflected the year’s preoccupations: industrial AI, workforce transition, US manufacturing competitiveness, and the safety standards needed to make all of it deployable. Automate has become an unusually faithful annual barometer of where industrial automation actually is, as distinct from where it claims to be; this year the gap between the two narrowed visibly.
Source: Business Wire
MIT Researchers Build Low-Power Chip Enabling Tiny Robots to Map Complex Environments
On June 23, MIT News described a new chip designed to let centimeter-scale autonomous robots build detailed 3D maps of their surroundings using roughly the power consumed by a single LED. The work pairs an algorithmic redesign of standard occupancy mapping with dedicated silicon, eliminating the memory and compute bottlenecks that have kept obstacle-aware navigation out of the smallest UAV class.
The motivating example the team uses is a small drone navigating the inside of an industrial HVAC system to inspect for gas leaks — a job too dangerous and too cramped for inspectors but historically too power-hungry for an autonomous flier of the size needed to fit. By doing the mapping work in custom hardware, the chip lets a battery-limited robot construct, update and use its own 3D map in real time.
Foundational compute work like this tends to be underappreciated in robotics coverage, but it determines the boundary of what is buildable. Once efficient on-board 3D mapping is possible at gram-scale, a long list of applications — pipe inspection, post-disaster search, indoor agriculture monitoring, large-warehouse inventory — moves from research demo to engineering project. The pattern mirrors the earlier curve where smartphone-class compute made low-cost consumer drones possible at all.
Source: MIT News
Surgical Robotics Industry Awards 2026 Recognize a Global Slate of Winners
Announced June 24, the fourth annual Surgical Robotics Technology Industry Awards named winners across categories including Outstanding Company, Groundbreaking Technology, Innovative Start-up and Software Innovation. SS Innovations International became the first Indian company to win Outstanding Company; the Groundbreaking Technology award went to Sentante, Software Innovation to Moon Surgical, and Innovative Start-up to Microbot Medical.
The geographic spread of the winners is worth pausing on. Surgical robotics has been dominated for two decades by a small group of US-headquartered platform vendors, with Intuitive Surgical’s da Vinci system holding overwhelming market share. A field where India, Lithuania (Sentante’s home), France (Moon Surgical) and an Israeli-American start-up like Microbot Medical are all simultaneously recognized for first-rank innovation is a noticeably more open one than it was even three years ago.
The awards are peer-voted by surgical-robotics professionals, which gives them more weight than vendor-funded prizes. The 2026 slate, taken together, suggests that the next generation of surgical robots will not be a single dominant arm-and-tower platform but a more modular field of single-purpose autonomous tools, software companies and minimally invasive specialist devices. That is a healthier outcome for hospitals and patients than the single-vendor pattern that preceded it.
Source: Surgical Robotics Technology
Hamlyn Symposium on Medical Robotics Opens with LMIC Focus
The Hamlyn Symposium on Medical Robotics, hosted by the Hamlyn Centre for Robotic Surgery at Imperial College London, ran from June 23 through June 26 at the Royal Geographical Society. This year’s theme — “Medical Robotics: the Complex Journey from Ideation to Value Creation” — was matched by a deliberate programmatic emphasis on frugal innovation and deployment in low- and middle-income healthcare settings.
The accompanying Surgical Robot Challenge, an international competition that has spotlighted next-generation surgical robot designs for over a decade, explicitly asked 2026 entrants to demonstrate practical impact in LMICs. The competition rewards teams whose 2-minute video submissions show technologies designed around real cost, supply-chain and operator-training constraints, rather than the marginal improvements that dominate the established surgical-robot pipeline.
For a field whose flagship platforms cost on the order of $1.5 million per unit and require dedicated infrastructure, the LMIC framing is not a token gesture. The vast majority of the global surgical demand that goes unmet is in countries that cannot afford the current generation of platforms. A research community openly orienting prize incentives toward affordable, simpler systems is how that gap eventually starts to close — and the Hamlyn Centre is one of the few venues with both the convening power and the technical credibility to set that agenda.