Intrinsic open-sources core robotics stack

Recap

Source: The Robot Report, report of Sept. 22, 2026

According to the report, Intrinsic, an AI robotics division now part of Google, has made its Intrinsic Core software stack publicly available. The company describes this release as a collection of pre-configured, ROS-compatible building blocks intended to streamline the creation of robotic applications. By providing these foundational components, Intrinsic aims to help developers avoid the extensive coding required to build such infrastructure from the ground up. The software is hosted on the company's GitHub and distributed under the Apache 2.0 license.

Brian Gerkey, CTO of Intrinsic, emphasized that this is a long-term commitment to the community rather than a one-time code drop. He noted that the team extracted capabilities from real-world manufacturing deployments to create a durable software environment. The core of the release is Intrinsic Control, a hardware-agnostic framework that enables real-time, sensor-based adaptation. This allows robots to adjust their behavior mid-trajectory and swap out arms or sensors without rewriting drivers.

The package also integrates several specific tools: pose estimation using NVIDIA’s FoundationPose for 6D object tracking, automated motion planning to generate collision-free paths, and grasp planning for dynamic part handling. Additionally, it includes Gazebo-powered simulation services, automated camera calibration, and pre-tuned ROS drivers for supported hardware. To help users get started, Intrinsic also released the Open Machine Tending Solution (OMTS), a ready-made reference design for CNC machine tending that supports robots from FANUC and Universal Robots.

Context

Intrinsic has been a significant player in industrial robotics, particularly in machine tending and material handling, before joining Google. The Robot Operating System (ROS) has long been the standard middleware for robotics research and development, but it often requires developers to assemble complex infrastructure for production-grade reliability. This release positions Intrinsic Core as a higher-level abstraction layer on top of ROS, similar to how containerization tools like Kubernetes simplify deployment in software engineering. The inclusion of NVIDIA’s FoundationPose highlights the growing trend of integrating large-scale AI foundation models into real-time robotic control loops.

Robot's take

This move signals a shift toward more standardized, production-ready robotics platforms. By open-sourcing the core stack, Intrinsic lowers the barrier to entry for developers who want to build reliable manipulation systems without reinventing the wheel. However, the platform is currently limited to robotic arms and does not support mobile robots or humanoids. The success of Intrinsic Core will depend on whether the community adopts it as a standard layer, potentially creating a new tier of robotics development between raw ROS and proprietary vendor stacks. It is not yet clear how this will interact with other emerging AI-robotics frameworks, but it sets a strong precedent for open-source industrial robotics.