RealSense D585 Pro — AI-Native Stereo Depth Camera for Physical AI and Robotics

Specifications

Overview

The RealSense D585 Pro is a rebuild rather than a revision. Instead of adding features to the existing depth module, RealSense moved to a Gen 5 SoC with a new depth engine, ISP, DSP and quad-core ARM processor, plus new dual projectors and new high-resolution sensors. The result is an AI-native depth camera designed for physical AI systems — autonomous mobile robots, humanoid platforms and vision-language-action pipelines.

The headline numbers are a 120 x 100 degree field of view at 60 fps, depth output up to 1280 x 960 at up to 90 fps, a minimum depth under 12 cm and reliable operation from 12 cm out past 10 m. RealSense claims 2x better depth quality across the entire field of view versus its current generation, 2x noise reduction with finer detail preservation, and 2.5x better close-range performance than competing cameras — the close range that matters when a robot arm or AMR is docking or manipulating.

What sets the D585 Pro apart for deployment is what is included by default. IP65 ingress protection ships on every unit with no premium SKU, connectivity spans GMSL2 and USB-C with hardware sync for GPU and x86 hosts, and built-in IR filters support both bright outdoor sunlight and low-light interiors. Dual RGB enables simultaneous 30 fps colour and 30 fps depth streams at up to 1280 x 960, merged on-camera so there is no host CPU cost — a configuration aimed at humanoids, inspection pipelines and digital twins. On-camera person detection runs on the device, an IMU is integrated, and a growing SDK delivers compression, VIO and occupancy grid capabilities to existing units over time. RealSense Perception Studio provides continuous-release early access to next-generation perception features, and the D585 has been announced for release in early 2027.

Key Benefits

Depth quality as the differentiator — 2x better depth across the full field of view and 2.5x better close range than competing cameras, with a 120 x 100 degree FOV and 12 cm to 10 m+ range. Deployment features included, not optional: IP65 on every unit, GMSL2 and USB-C with hardware sync, built-in IR filters and an IMU. On-camera AI and dual RGB with merged 30 fps colour plus depth at no host CPU cost.

Applications

AMR and AGV navigation, humanoid robot perception, visual SLAM and whole-scene awareness, inspection pipelines and digital twins, vision-language-action model input, warehouse and factory automation, and outdoor mobile robotics requiring IP65 and IR filtering.

Related Products