RealSense unveils AI-native D585 Pro depth camera for robots

NVIDIA Corp. — realSense is unveiling the new D585 Pro AI-native depth camera at Automate 2026 in booth 12036. The RealSense D585 Pro combines depth sensing, edge AI acceleration and a software-defined platform that the company said is designed to improve over time through SDK-delivered capabilities.

The RealSense D585 Pro is expected to begin shipping in Q1 2027. Designed for humanoids, autonomous mobile robots (AMRs), collaborative robot arms, industrial robotics and inspection systems, the D585 combines a sub-15cm minimum range at full resolution, a 120×100° field of view (FOV) at 60 FPS, IP65 protection as standard and an AI inference engine that runs directly on the camera at the edge. RealSense said the D585 Pro is powered by a proprietary Gen 5 system-on-chip (SoC) and delivers more than 2x better depth quality than the previous generation of RealSense cameras, enabling more precise navigation, manipulation, inspection and human-robot interaction. With on-device AI processing, the camera delivers full depth and image processing pipeline, reducing reliance on host computing resources while opening the door to a new generation of intelligent robotic applications. “The D585 Pro is not just an update to what came before; it’s the actualization of the Visual Cortex of Physical AI,” said Nadav Orbach, RealSense CEO. “For the first time, developers can deploy a single depth camera that operates from under 15 centimeters to more than 10 meters, indoors or outdoors, while gaining new capabilities over time through software updates rather than hardware replacement. The early response from leading robotics companies has reinforced our belief that the industry is ready for a new generation of perception platforms that combine depth sensing, AI and software-defined capabilities in a single device.” The D585 Pro introduces a new hardware architecture built around a custom Gen 5 SoC featuring a depth engine, image signal processor (ISP), digital signal processor (DSP), dedicated AI accelerators, and a quad-core ARM processor, combined with dual IR projectors and high-resolution sensors. Key specifications at launch include: 120×100° FOV, optimized for Visual SLAM, navigation and whole-scene awareness. Dual RGB, delivers depth and imaging for VLAs and world models, and enables a path from tele-operation to fully autonomous robots 2x better depth quality, across the entire FOV, with finer detail visibility, 2x noise reduction and dramatically reduced temporal flicker. Sub-15cm minimum range at full resolution, 2.5x better than the nearest competitor, it enables close-range applications, including cobot arm inspection, bin picking and shelf scanning that traditional depth cameras cannot support. 60 FPS at 1280×960, double the frame rate of 30 FPS class competitors, sustaining full-resolution capture on fast-moving robots and high-speed conveyors. 10m+ operating range at optimal accuracy, covering the full navigation envelope for AMR fleets in warehouse and factory environments. IP65 as standard, on every unit, with no premium SKU required. GMSL2 + USB-C connectivity, with hardware sync support for GPU and x86 deployments. Built-in IR filters, for reliable operation in bright outdoor sunlight and low-light indoor environments. The D585 Pro ships with on-camera enhanced depth processing and person detection in beta at launch, running entirely on the Gen 5 SoC without requiring host compute.