Stream Motion Meets the Real Factory: How Dynamic Applications are Becoming Practical

For decades, General Assembly has been the last room in the factory where robots still hesitate. Stamping, body-in-white, paint, automated long ago. But the moment a part starts moving on a continuous conveyor and tolerances tighten to a few millimeters, the traditional playbook collapses. Fixture it, index it, stop the line, or skip it. Most automakers picked some combination of all four.

That trade-off is starting to disappear. With Stream Motion (STMO), FANUC's real-time external motion option, robots no longer have to wait for the part to come to them. And with partners like Inbolt building perception and intelligence on top of it, the moving line is shifting from an exception to default operating condition.

What Stream Motion Actually Does

Stream Motion is, in technical terms, a low-level interface that hands trajectory authority to an external system. The robot follows the external trajectory while the controller continues to enforce velocity, acceleration and jerk limits, keeping motion inside safe envelopes.

In plain terms, the path the robot follows can be calculated by something that knows things the robot doesn't: where the part actually is, where it's heading, and how it's shifting on a conveyor that never stops. It enables sensor driven robot servo-control, the major unlock to more intelligent robots.

The option is available across FANUC's controller lineup: the R-30iB (software 8.x), the R-30iB Plus (the default on CRX for the past several years, software 9.x), and the newest R-50iA (software 10.x), which adds higher-frequency streaming and native Python and ROS 2 support for the next wave of AI-driven workloads. Together, that covers all CRX cobots and the majority of industrial yellow robots already on factory floors today.

Where Vision Comes In

Real-time control is only half of the problem. To correct a robot's path on the fly, something has to see the part quickly, accurately, and without custom lighting or fixturing. That is where Inbolt has built one of the more interesting applications running on top of STMO.

Inbolt mounts a compact 3D camera on the robot wrist, runs its AI vision model to localize the part in the part's own reference frame, and streams corrected joint commands into the controller through Stream Motion. The robot tracks parts moving down the line and executes the planned trajectory (screw insertion, bolt rundown, glue application, de-racking) without indexing, halting, or re-teaching.

This is not confined to lab demos. Inbolt's solution is already deployed on leading automotive lines, with multiple major projects in production using CRX cobots for dynamic applications. The system handles part variability, imperfect environments, and the crowded General Assembly stations where conventional automation either doesn't fit or doesn't pay back. Inbolt's solution runs today on the R-30iB Plus controller that ships with CRX, with industrial-yellow and R-50iA support landing next.

Why This Matters on the Plant Floor

Three things happen when STMO and vision are combined on a real production line.

Cycle time stops being a tax. Conventional moving-line workarounds either stop-and-clamp or restrict themselves to slow line speeds. With external trajectory correction at 125 Hz, the robot keeps pace with the line at its native speed, and the precision task finishes inside the same takt time.

Infrastructure cost drops. No indexing tables, no part presentation rigs, no bespoke fixtures per SKU. A single station can run dozens of variants, in Inbolt's case, over 80 part models, without re-tooling between runs.

Deployment moves from weeks to days. Programs are built directly from CAD inside Inbolt Studio, the vision model trains against synthetic data, and the robot executes the planned path the moment a real part arrives. The trajectory the engineer designed in the digital twin is the same trajectory that runs on the floor.

The combined effect is what FANUC and the market has been calling Physical AI: perception, motion, and execution working as one system, on the robots manufacturers already own.

See it Run at Automate 2026

If you want to see Stream Motion in action, Automate in Chicago, June 22–25, is the place. FANUC's Booth #1401 will run a CRX-20iA/L performing real-time bolt tightening on a moving engine block, with Inbolt's robot intelligence and NVIDIA-powered processing in the loop. The Inbolt Booth #1675 will run three additional CRX demonstrations: dynamic dispensing on a moving engine, real-time workpiece tracking, and de-racking — each programmed from CAD and executed live.

For decades, "the line doesn't stop" was the constraint that kept high-precision automation out of General Assembly. With Stream Motion and the right perception layer on top, it is starting to be the reason to put it in.