NVIDIA Robotics

Today at #ROSCon2026 in Toronto, we’re taking a major step forward in our mission to democratize access to intelligent robotics. We’re excited to introduce Intrinsic Core™ - open-sourcing foundational capabilities from the Intrinsic platform so developers worldwide can build adaptive, physical AI applications more efficiently. Free and released under an Apache 2.0 license, Intrinsic Core works with the Open Robotics Suite, is compatible with ROS and can run locally on your machine. It brings the same building blocks we use for real-world manufacturing deployments directly to the open robotics ecosystem. Some of the capabilities and services include: 🔹 Intrinsic control: A hardware-agnostic, real-time framework that delivers fast, sensor-based control. 🔹 Pose estimation: Leverages NVIDIA FoundationPose® to enable robots to dynamically detect parts and their location with high accuracy. 🔹 Motion planning: Automates robot path planning so you don’t have to manually program joint by joint, in a position-controlled way. 🔹 Simulation services: Powered by Gazebo, it provides a straightforward way to visually test and troubleshoot robotic solutions that update as you build. To get started quickly with Intrinsic Core, we’ve built a pre-configured Open Machine Tending Solution. This ready-made reference solution for intelligent CNC machine tending gives developers a starting point to customize their own solution and get to grips with Intrinsic Core capabilities and services. Anything you build with Intrinsic Core will work with other Intrinsic offerings, including our enterprise services - no code refactoring or rewrites needed. Our goal is to make it much easier to go from pilot to industrial grade production.  Learn more and get started today: 🦾 Download Intrinsic Core on GitHub now: https://lnkd.in/gfd5eg4r 📖 Read the full announcement: https://lnkd.in/gq-jzkyq 🌐 Join our Developer Community: https://lnkd.in/gUTYStCi Thank you to our amazing partners: NVIDIA, Universal Robots, FANUC America Corporation, SCHUNK - Hand in hand for tomorrow, ATI Industrial Automation, Robotiq, Basler AG, Open Robotics

ROS Lyrical Luth gains Vendor-Neutral Accelerated Memory Transport from NVIDIA Robotics. These buffers are ROS-native and designed from the ground up for compatibility. The feature is entirely vendor-neutral, and is intended to support multiple hardware vendors, memory types, middleware implementations, and accelerated libraries. https://lnkd.in/exZi8Epp

From just 32 input images to real-time exploration of NVIDIA’s Voyager headquarters. Atlas is World Labs’ next-generation multimodal world model, trained on NVIDIA Blackwell GPUs. It uses those frames as 3D spatial context to reconstruct the building’s complex layout, letting you explore interactively with pixel-perfect camera control as it generates new views in real time. For design and visualization teams, this demonstrates a new way to explore layouts, evaluate viewpoints, and move beyond the original reference frames. For robotics teams, it opens possibilities for reconstructing real-world environments and generating new viewpoints for simulation. See our real-time version of Atlas in action in the highlight reel below - and take a look around.