Open Source RoboticsRobots, software & learning

Intrinsic Open Machine Tending Solution

Software

Configurable CNC machine-tending reference application using Intrinsic Core, behavior trees, perception and machine I/O. [1]

Installation & first result →Official websiteGetting started

For
Robotics developers evaluating local industrial manipulation software [1]
Hardware
Documented lab_bb_01: UR3e / CAW enclosure without CNC or vise [5] [6] [7]
Source rights
Named licenses are documented by part. Review licenses

Source researched · Not tested by us
Sources checked · Research details

Licenses & related hardware
Ecosystem
ROS 2, Gazebo

Requirements · Research details and dates

On this page

Overview

OMTS supplies the application layer for a CNC tending cycle: pick stock, load the machine, coordinate the machining handshake, unload and return the workpiece to infeed. It runs on Intrinsic Core and organizes those steps as a behavior tree. This profile covers software release 20260922.0, not a purchasable or fully open hardware cell. [1] [10]

Is this for you?

Workflows
  • Deploy the OMTS scene in simulation
  • Inspect belief-world state in RViz or physics state in Gazebo
  • Configure and run the OMTS behavior tree after scene and pose-estimator setup

[3] [4] [5]

Known limitations
  • No installation or robot execution reproduced by this publication
  • ARM hosts are unsupported in the pinned guide
  • Perception requires dedicated NVIDIA compute and separately licensed model weights
  • A configured software cell does not establish a complete physical-cell BOM, delivered cost or commissioning result

Combination of explicit upstream requirements and stated limits of this source review.

[3] [5] · Editorial assessment

What you need

Operating system
  • Ubuntu 26.04 for the linked installation guide

Core README lists broader versions; compatibility beyond this guide has not been reconciled.

[3]

Dependencies
  • Git LFS
  • GitHub CLI
  • k3s/containerd
  • Bazelisk / Bazel 8.8.0
  • ROS 2 Lyrical for the documented RViz route

[3] [4] [9]

Host computer

x86-64 host; Intel required by the guide for real-time hardware control, AMD permitted for simulation; ARM unsupported.

[3]

Compute for the workload
  • Guide recommends 6 cores / 12 threads at 4.9 GHz or higher
  • 32 GiB RAM minimum; 64 GiB recommended; smaller setups need constrained builds
  • 1 TB NVMe SSD with at least 100 GB free
  • Integrated graphics for basic simulation; NVIDIA RTX 3060/4060 or higher for ML/vision and full OMTS simulation with perception
  • 2–3 Gigabit Ethernet ports recommended

[3]

Installation & first result

Open the upstream getting-started guide [3] · Project documentation [1]

  • Prepare the guide’s Ubuntu 26.04 x86-64 host and fetch both matching releases with LFS
  • Install k3s, deploy intrinsic-base and install inctl from release 20260922.0
  • Configure NVIDIA GPU support when using perception
  • Install Bazelisk and follow the documented LLVM/libxml compatibility step
  • Build //:omts_solution with --config=lab_bb_01 and --operation_mode=sim
  • Use the documented ROS 2 Lyrical/Zenoh RViz route or Gazebo to inspect the deployed scene

[3] [4]

Installation

Follow the pinned Core guide: install Git LFS/GitHub CLI, fetch release 20260922.0 of Core and OMTS, configure k3s, deploy the matching runtime archive and inctl, then build the simulation with Bazelisk. Configure NVIDIA support for perception.

[3]

Hardware, models & data

Supported hardware
  • Documented lab_bb_01: UR3e / CAW enclosure without CNC or vise; configuration names Robotiq Hand-E and Orbbec camera
  • Default OMTS: UR5e / CNC enclosure / Schunk vise
  • Hardware adapters also document digital-I/O gripper and CNC machine interfaces

Specific source configuration and adapter evidence; not a claim of universal device compatibility or our hardware testing.

[5] [6] [7]

Interfaces
  • ROS 2 interoperability
  • inctl CLI
  • Local deployment endpoint on port 17080

[1] [3] [4]

Inputs and outputs

OMTS consumes a cell configuration, camera/perception results and machine I/O state; its behavior tree coordinates robot motion, gripping, CNC loading, machining handshake, unloading and return to infeed.

[5] [6] [7]

Project notes

Inspect the example cell before choosing devices

The README distinguishes the default OMTS cell (UR5e, CNC enclosure and Schunk vise) from lab_bb_01 (UR3e and CAW enclosure, without CNC or vise). That lab scene is useful for initial simulation, but it is not the complete CNC example. Its configuration names a UR robot module, a Robotiq Hand-E finger joint, an Orbbec camera and perception-based infeed. Its default cycle count is one. The adapter documentation also exposes digital-I/O gripper and CNC interfaces. These are concrete starting points for integration, not evidence that every arm, camera or CNC machine works without additional drivers or configuration. [1] [6] [7]

Door, vise and cycle signals belong to the application configuration. The hardware layer updates the software's scene representation alongside machine commands so planning can reflect door and vise positions. Source code describing that coordination does not establish that a particular physical installation has been commissioned or tested. [6]

Deploy simulation, then prepare the application

Use the linked Core installation guide with Ubuntu 26.04 on x86-64. It recommends at least 32 GiB RAM and 100 GB free storage; the full perception-enabled simulation requires dedicated NVIDIA graphics. Both repositories use release 20260922.0, and OMTS pins the exact Core commit in its dependency file. Install Git LFS before fetching model assets. [3] [8]

Deploy //:omts_solution with the documented lab_bb_01 configuration and --operation_mode=sim. Deployment is not yet the complete tending cycle: the OMTS README then applies scene updates, registers the workpiece pose estimator, and runs //src:omts_app with the matching configuration file. RViz and Gazebo provide different views of the deployed scene. [1] [3] [4]

What we checked

Source-researched release profile. We have not installed this release, run its tests, commissioned a cell or measured performance. No affiliation or endorsement claimed.

Updates

Upstream last updated
Sep 22, 2026 [10]
Latest release
Sep 22, 2026 [10]
Profile updated
Oct 4, 2026
Sources last checked
Sep 30, 2026
Next review
Oct 30, 2026
Reviewed revision
20260922.0 / 8253cdfdd173da9d5b7c9bb7b8cee817f66902ae · software repository

Source files and licenses

Open software profile; external model weights, dependencies and physical devices have separate scope. No complete open hardware or tested deployment claim.

LayerAvailability / license
softwareavailable · Apache-2.0
Repository material under its root license; third-party artifacts and marks retain separate terms. [1] [2]
model weightsavailable · NVIDIA Open Model License (upstream OMTS declaration)
External FoundationPose weights downloaded by the build; absent from the reviewed repository and not covered by its Apache-2.0 license. This review records the upstream declaration, not a legal interpretation or verified download. [5]
documentationavailable · Apache-2.0
Repository material under its root license; third-party artifacts and marks retain separate terms. [1] [2]

No training dataset is being offered or reviewed in this runtime/application profile.

Keep model licensing and cell cost separate

OMTS's code carries Apache-2.0. Its README explicitly excludes the FoundationPose ONNX weights from that license: the build retrieves them from NVIDIA under separate NVIDIA Open Model License terms. Preserve that distinction when deciding which artifacts to redistribute. [2] [5]

The software example does not give us a verified, delivered price for a complete robot, gripper, camera, host and CNC integration. We have not built this release, run its tests or operated the reference cell; those remain the next steps for a reproduction effort.

References

Each source links to the exact revision we read. “Used for” shows where this page relies on it.

Project sources

Videos

  • S011 Intrinsic (YouTube channel)
    The project’s own YouTube channel, found by searching YouTube and checked against the channel page. Not reviewed by us for accuracy.
    Revision, date & where cited

    YouTube channel Intrinsic · youtube.com · Accessed Oct 4, 2026

Correction guidance · Discussion is not enabled · Profile format 0.6.0-draft.1

Change history

2026-09-30 — Initial source-researched profile of release 20260922.0; installation scope and external weight terms distinguished.

2026-10-04 — Added 1 YouTube reference(s) (S011): official channel or videos linked from the pinned upstream sources. Existing claims, source-check dates and evidence are unchanged.