RealSense D585 Gives Robots Smarter, Faster 3D Vision
The new RealSense D585 depth camera features on-chip AI and a 20m range to give robots a plug-and-play 3D view of the world.
When a robot, drone, or other autonomous system needs a clear understanding of its surroundings, developers often choose to equip it with a depth camera. These sensors make it possible to measure the distance to nearby objects and build a three-dimensional map of the environment, helping machines navigate, avoid obstacles, and interact with the world around them. The downside is that adding a depth camera is only the beginning — engineers will still need to develop the algorithms and control systems that are powered by the data these cameras produce.
The latest release from RealSense — the D585 — was built to simplify computer vision applications. Not only does it provide better image quality than past models, but it also does a lot of image processing on-device. That frees developers up to create without worrying about every single little detail.
Processing at the edge
At the core of the D585 is a new RealSense V5 system-on-chip that combines a depth engine, image signal processor, DSP, AI acceleration hardware, and a quad-core Arm processor into a single package. Instead of streaming raw data to a host computer and expecting it to handle everything, the camera can perform many tasks internally. RealSense describes the device as an AI-native, software-defined vision platform that will continue to gain new capabilities through SDK updates after deployment.
The stereo depth camera features dual infrared projectors, high-resolution sensors, global shutter imaging, and integrated IR filters that allow it to operate both indoors and outdoors. A 120° by 100° field of view gives robots a broad view of their surroundings, while depth output resolutions reach up to 1280 × 960 pixels. Full-resolution operation is supported at up to 60 frames per second, with depth frame rates reaching as high as 90 frames per second.
RealSense says the D585 can measure depth at distances as short as 10 cm at full resolution, making it suitable for applications like robotic manipulation, inspection systems, and collaborative robot arms working near objects. On the opposite end of the spectrum, the company lists an ideal operating range beyond 10 meters and a maximum range exceeding 20 meters, allowing the same camera to support mobile robots in large warehouses and industrial facilities.
Reducing system complexity
The D585 also introduces features designed to reduce system complexity. A Dual RGB mode can output synchronized RGB and depth streams simultaneously while handling stream alignment directly on the camera. RealSense has also developed a Depth Compression feature that reduces bandwidth requirements by approximately 75 percent, which could be useful when streaming high-frame-rate depth data across bandwidth-constrained links.
Additional capabilities are planned through the RealSense SDK ecosystem. Person detection will be available in beta at launch, while future updates are expected to add features such as visual-inertial odometry, occupancy grid generation, auto-calibration, face detection, and enhanced depth processing.
RealSense expects the camera to begin shipping in the first quarter of 2027. Pricing has not yet been announced.