RealSense unveils D585 Pro AI depth camera

A depth camera that runs AI on board is reshaping how robots see. Unveiled at Automate 2026, RealSense's D585 Pro combines depth sensing, edge AI acceleration and a software-defined platform, with shipping planned for Q1 2027.

The company reports that the D585 Pro is powered by a proprietary Gen 5 system-on-chip and delivers more than 2x better depth quality than its predecessor, enabling tighter navigation, grasping and inspection tasks. It is designed for humanoids, autonomous mobile robots, collaborative robot arms, industrial robotics and inspection systems, a portfolio RealSense says will benefit from on-device AI processing that reduces the need for heavy host computing. The camera's edge inference engine runs directly on the device, delivering a full depth and image processing pipeline at the edge and minimizing data backhaul to central compute resources. RealSense CEO Nadav Orbach framed the product as more than a spec upgrade, calling it "the actualization of the Visual Cortex of Physical AI." He said the D585 Pro can operate from under 15 centimeters up to beyond 10 meters, indoors or outdoors, and that software updates will unlock new capabilities without hardware swaps. "For the first time, developers can deploy a single depth camera that operates across a broad range while gaining new capabilities over time through software updates rather than hardware replacement," the company reports, noting early responses from robotics players reinforcing the market's readiness for software-defined perception.

The hardware details anchor practical expectations for engineers. The D585 Pro carries a sub-15 cm minimum range at full resolution and a wide 120 by 100 degree field of view at 60 frames per second, wrapped in IP65 protection as standard. The combination supports precise navigation, manipulation, inspection and human-robot interaction at close range, as well as extended outdoor or indoor perception at longer distances. RealSense positions the device as a versatile core for humanoids and AMRs, enabling perceptual tasks to run locally while still benefiting from cloud or on-platform analytics via software updates. The on-camera AI processing helps apps stay responsive even when connectivity to a host computer is limited or intermittent, a meaningful advantage in robot arms and mobile platforms that must react in real time.

From a practitioner's perspective, several constraints and opportunities stand out. First, edge AI at the camera level promises lower latency and reduced data bandwidth, which is critical for real-time navigation and manipulation. But developers must balance the onboard compute budget with power and thermal limits typical of a compact edge sensor, especially in environments with variable lighting and clutter. Second, a software-defined approach raises the value of a stable SDK and robust update cycles; continuous improvements are possible, yet teams need solid testing, backward compatibility and clear migration paths to avoid destabilizing safety-critical workflows. Third, the wide operational range (sub-15 cm to over 10 m) expands use cases but demands careful sensor fusion and calibration to maintain depth accuracy across tasks, from fine-grained grasping to broader mapping. Fourth, the IP65 rating supports rugged use, but practitioners should anticipate the ongoing need to validate perception stacks across different robot platforms and payload requirements as software capabilities evolve.

Overall, the D585 Pro marks a meaningful step toward a software-defined perception stack for robots, combining a refined depth sensor, edge AI acceleration and a platform designed to evolve without hardware swaps. If the early feedback from the field translates into solid developer tooling and reliable updates, this could shorten integration cycles for humanoids and autonomous systems that must adapt to new tasks without new hardware.