At Chicago's Automate 2026 trade show, Orbbec and Robbyant (Ant Group's robotics arm) showed AI enhanced 3D cameras aimed at the transparent and reflective surfaces that still fool factory robots. Every performance figure is vendor reported and unreviewed.
When a robot reaches for a clear plastic bottle on a packing line, its 3D camera often sees right through it. The same goes for polished metal brackets and featureless cardboard stacks. The depth sensors that guide most industrial robots cannot tell where the object ends and the background begins, so the arm misjudges the grasp, drops the part, or stops the line. That failure mode is the actual problem Orbbec and Robbyant, the robotics arm of Chinese fintech giant Ant Group, said they have begun to solve with AI-enhanced 3D vision shown at this week's Automate 2026 trade show in Chicago.
The mechanism behind the failure is structural. Standard industrial 3D cameras, which use structured light or time-of-flight sensors to build a depth map of the scene, work by projecting a pattern of infrared dots or measuring how long light takes to bounce back. Transparent materials let the projected light pass through, so the camera reads the surface behind the bottle rather than the bottle itself. Reflective metal scatters the pattern into noise. A conveyor full of identical cardboard flaps gives the sensor nothing distinctive to lock onto. Engineers have worked around these cases for years with extra lighting, fixture jigs, and hand-tuned thresholds, but each fix adds cost and slows down the line.
The pitch from Orbbec and Robbyant is that a vision-language-action model, a class of multimodal AI that fuses image understanding with action planning, can learn to read depth maps even when the underlying geometry is missing or corrupted. Orbbec's Gemini 330 series cameras feed structured-light and time-of-flight data into LingBot-Depth, an AI model co-developed with Robbyant, which the companies said fills in the gaps and recovers usable 3D coordinates. Processing happens on an NVIDIA Jetson Orin edge-AI module mounted on the camera, so inference stays on the factory floor rather than round-tripping to a cloud server. Trade press covering the launch framed the system as targeting "AI-enhanced 3D vision that addresses industrial blind spots."
The Ant Group connection is the part that lifts this above a routine product launch. Ant Group is best known in Western markets for the Alipay mobile payments network and a years-long IPO saga in China, not factory automation. Its robotics arm Robbyant has been quietly building a vision-language-action AI stack, and the Orbbec partnership is its first public industrial test case. If the approach holds up on transparent bottles and shiny brackets, it would mark a credible bid by a fintech-affiliated AI lab into edge industrial perception, a market long dominated by incumbents such as Keyence, Cognex, SICK, and Basler.
Three caveats belong next to the promise. Every performance figure released so far, including success rates on transparent vessel handling, reflective metal alignment, and carton edge tracking, comes from Orbbec's own demonstration materials and Robbyant's published claims, as relayed in the trade press coverage. None of it has been independently benchmarked, and the announcements were timed for a trade show floor rather than a peer-reviewed venue. The LingBot-Depth model is described as a co-development with Robbyant, but the commercial terms, including whether the arrangement is exclusive, what regions it covers, and how revenue is split, have not been disclosed in any of the company filings or partner statements surfaced by trade outlets. And the broader market question, whether factory automation buyers will trust a payments company's AI lab to handle a vision-critical cell on their line, remains open.
The watch items from here are concrete. Independent test data on transparent and reflective surfaces, ideally from a third-party integrator rather than a vendor demo, is the single piece of evidence that would move this from announcement to credibility. Any disclosure of the Ant Group side's commercial intent for Robbyant, including whether Orbbec is an exclusive camera partner or one of several, would clarify whether this is a one-off reference design or the opening move in a broader industrial strategy. Booth traffic and follow-on order announcements from systems integrators at Automate 2026 will be the earliest real signal of buyer demand. Until then, what Orbbec and Robbyant showed in Chicago is a credible mechanism, a notable new entrant, and a set of vendor-reported numbers that the industry has not yet pressure-tested.