This year’s ROSCon conference is taking place in Singapore, uniting the global robotics developer community centered around the Robot Operating System (ROS), the most widely used open framework for robot development. At the event, running through Wednesday, October 29, NVIDIA unveiled new collaborations with industry partners and the Open Source Robotics Alliance (OSRA), along with key software advancements aimed at accelerating robotics innovation. NVIDIA is supporting OSRA’s newly formed Physical AI Special Interest Group, which focuses on real-time robot control, accelerated AI processing, and enhanced tools for enabling autonomous behavior in robots. This initiative is part of a broader mission to establish ROS 2 as the go-to open, high-performance framework for real-world robotic applications. A major technical contribution from NVIDIA is the integration of GPU-aware abstractions directly into ROS 2. These abstractions allow ROS 2 to recognize and efficiently manage diverse computing hardware—including CPUs, integrated GPUs, and discrete GPUs—ensuring consistent, high-speed performance and future-proofing the ecosystem against rapid hardware evolution. To help developers improve robot performance and reliability, NVIDIA is open-sourcing Greenwave Monitor, a tool that quickly identifies performance bottlenecks in robotic systems, significantly speeding up development and debugging cycles. NVIDIA also announced the availability of NVIDIA Isaac ROS 4.0, a new suite of ROS-compatible, GPU-accelerated libraries and AI models. Built for the NVIDIA Jetson Thor platform, Isaac ROS 4.0 enables developers to deploy physical AI and robotics applications with access to CUDA-accelerated libraries, AI models, and end-to-end workflows for robot manipulation and mobility. These advancements are already empowering developers and partners worldwide to train, simulate, and deploy next-generation robots using NVIDIA’s AI, accelerated computing, and open-source technologies. AgileX Robotics leverages NVIDIA Jetson modules for AI-driven autonomy and vision in its mobile robots, while also using NVIDIA Isaac Sim—an open-source robotic simulation framework built on NVIDIA Omniverse—for high-fidelity simulation and testing. Canonical is simplifying robot development by introducing a fully open observability stack for ROS 2 devices on Ubuntu, now available on the NVIDIA Jetson AGX Thor platform, supporting robotics and edge computing use cases. Ekumen Labs has integrated NVIDIA Isaac Sim into its workflows, enabling photorealistic simulations for system validation and generating synthetic data to train AI models. Intrinsic is incorporating NVIDIA Isaac foundation models and Omniverse simulation tools into its Flowstate platform, enabling advanced robot grasping, real-time digital twin visualization, and AI-driven automation for industrial robotics. KABAM Robotics’ Matrix robot uses NVIDIA Jetson Orin and the NVIDIA Triton Inference Server on ROS 2 Jazzy to deliver advanced security and facility management in complex outdoor environments. At ROSCon, Open Navigation will present a keynote by founder Steve Macenski titled “On Use Of Nav2 Route,” demonstrating advanced navigation capabilities for autonomous mobile robots using NVIDIA technologies, including Isaac Sim and SWAGGER. Robotec.ai and NVIDIA are collaborating on a new ROS simulation standard now integrated into Isaac Sim, streamlining cross-simulator development and enabling more robust, automated testing. ROBOTIS uses NVIDIA Jetson for on-board computing and Isaac Sim for simulation and validation. Its AI Worker, powered by the Isaac GR00T N1.5 model, delivers enhanced autonomy and scalable edge AI. Stereolabs’ ZED cameras and ZED SDK are now fully compatible with the NVIDIA Jetson Thor platform, supporting high-performance, multi-camera capture, low-latency perception, and real-time spatial AI vision for general-purpose robotics. From foundational open-source contributions to powerful simulation tools and production-ready hardware, NVIDIA continues to provide the ecosystem with the tools needed to build the future of physical AI.