Vision narrow
Detail and reach for part identification at the working distance of the arm.
Applications
Robotics
Arm guidance, pick-and-place, and bin picking — and the mobile robots that carry the parts between them — with metric depth measured where the machine acts.
Challenge
What makes this hard
A 2D camera tells the arm what an object looks like. It does not tell it how far away the object is, how it is tilted, or whether the bin behind it is empty. Every one of those is a metric question.
The usual answer is to bolt a depth sensor next to a color camera and reconcile the two afterward. That means two calibrations, two clocks, two coordinate frames, and integration work that recurs on every new cell. The industry cost of that reconciliation is what the platform removes.
Distance to the part, not an estimate
Pose and tilt at the gripper
A decision inside the machine cycle
Configuration
The module mix
ToF 4-VCSEL
Depth across the bin, 940 nm at 0.2–20 m with ±0.5% accuracy.
Ethernet body
One cable to the cell controller; the same body across a fleet of arms.
Outcome
What changes
01
One calibration
Color and depth arrive in the same frame from a factory-calibrated head, so the cell integrates one device.
02
On-device decisions
Inference runs on the edge stage in 10–20 ms, inside the motion cycle rather than after it.
03
Reusable across cells
A new cell changes the module mix, not the vision stack — the SDK and ROS 2 interface stay put.
Figures for your program come with the proposal.