Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

intrinsic-ai/intrinsic-core

Hardware & IoT

A local runtime, SDK, and real-time control framework for building industrial robotics applications on ROS 2.

755
95
C++
Apache-2.0
2 months old

What it is

Intrinsic Core is an open-source, hardware-agnostic runtime and SDK for industrial robotics. It packages a local execution environment, a real-time motion and hardware control engine, motion planning, 6-DoF perception, edge model inference, a digital twin, and ROS 2 interoperability into one stack. The target audience is robotics engineers who need to build machine-tending or similar industrial applications without writing and wiring every low-level component themselves.

The runtime is packaged as a native k3s containerized environment; the control module (ICON) combines kinematics solvers, trajectory interpolation, and a deterministic control loop behind a hardware abstraction layer, so arms, grippers, and fieldbus I/O can be swapped without driver rewrites. Perception standardizes camera and point-cloud interfaces and includes NVIDIA FoundationPose for 6-DoF pose estimation of un-fixtured 3D parts. An open Machine Tending reference design (intrinsic-omts) shows a real shop-floor use case with pre-configured assets and skills.

Getting started

Prerequisites from the README: Ubuntu 24.04 LTS (Ubuntu 26.04 LTS listed as supported; Ubuntu 22.04 LTS also supported) and ROS 2 Lyrical Luth. Follow the Getting Started guide for full installation and setup; the README itself does not list the installation commands. Tutorials for customizing physical layouts, visualizing the robot, and more are under developer_resources/learn/tutorials, and sdk-examples demonstrates the Solution Builder Library.

To extend the system, use the SDK interfaces and serialization helpers to build reusable skills, custom hardware assets, and execution nodes that integrate with the runtime and digital twin. API definitions live in Intrinsic Core's interface definitions, and hardware drivers and manifests in the hardware module.

Constraints

  • The runtime is Linux-oriented: installation is documented for Ubuntu 24.04/26.04, with Ubuntu 22.04 supported.
  • ROS 2 Lyrical Luth is required.
  • The real-time control loop and k3s packaging assume a local physical or simulated robot setup; deploying to real hardware requires compatible drivers or a hardware abstraction layer implementation.
  • Contributions require signing the Google Contributor License Agreement and following C++20 standards and the pull request process in CONTRIBUTING.md. Security issues are reported privately per SECURITY.md.
  • The project is Apache 2.0 licensed, and the README notes it is not an officially supported Google product. "Intrinsic" and "Intrinsic Core" are trademarks of Intrinsic Innovation LLC (TRADEMARK.md).

Related repositories

The ecosystem includes MoveIt 2 integration (intrinsic-moveit), a local inference service (intrinsic-inference), hardware modules (icon-hwm-controller), shared-memory IPC (icon-shared-memory), camera drivers (intrinsic-ros-camera-drivers), and a real-time heap-allocation guard (malloc-guard). The developer community hosts guides, a forum, and support.

HotRepo's take

Intrinsic Core entered HotRepo's C++ daily ranking on 2026-11-19 at position 5, with 34 stars gained and 754 total stars that day. The project is recent (created 2026-09-08), so that single snapshot is a launch-period signal rather than an established trend. What stands out is the scope of the shipped stack: a pre-configured k3s runtime, a deterministic control loop, FoundationPose-based perception, and an OMTS reference design in one repository, backed by a named industrial-robotics organization rather than a generic ROS wrapper. The practical caveat is that hardware support depends on the HAL and peripheral integrations, so the useful next check is whether the README's supported arms and grippers match the target cell.

Written by AI from the project's README, updated . The project's own docs are the source of truth.

You might also like

  1. google-deepmind/robotics-antigravity-pluginAgent skills that let Gemini-based coding agents build MuJoCo simulations and control real robot arms.Python109
  2. inclusionAI/IDMResearch code and inference toolkit for one- or two-step text-to-image generation via intrinsic distribution matching.Python161
  3. rerun-io/rerunOpen-source SDK and viewer for logging, visualizing, and querying multimodal robotics and sensor data over time.Rust11,710
  4. ros-navigation/navigation2A ROS 2 navigation framework that gives mobile robots mapping, localization, path planning and autonomous movement.C++4,944