UC-13 · Vision-guided robotics

Robot Guidance & Pick-and-Place

Parts that arrive in random orientations. Bins of mixed parts. Conveyors with parts that don't index. 2D and 3D vision guides the robot to the right part, in the right orientation, at the right pick angle — at line speed and without fixturing.

REC INSPXION X-12B LIVE INSPXION X-12B · ROBOT GUIDANCE 1,800 PICKS/HR 0.4% MISS LIVE InspXion X · 464 Y · 336 θ · 0° SPEC PROTOCOLS FANUC · ABB CYCLE 0.4–1.2 s ACCURACY ±0.2 mm 3D OPTIONAL PASS · PICKED PICKS 12,847 CYCLE 0.6 s MISSED 52
Why it matters

Fixturing parts for blind robotic picks is the slow, expensive way to do automation.

Fixturing adds tooling cost, changeover time, and a constraint on every new product. Vision-guided picking eliminates fixturing — the robot picks whatever the camera tells it is there, in whatever orientation. New products onboard in days, not months.

Two paths to the same outcome

Retrofit, or build it in from the start.

Retrofit · use what you have

If you have a robot doing fixtured picks.

Most existing robot cells use mechanical fixtures or bowl feeders to present parts in known positions. We add a 2D vision station upstream of the robot; the camera sends per-part position and orientation to the robot controller. Fixtures stay (as fallback / overflow) but most picks become vision-guided.

  • 2D vision overlay on existing robot cells
  • Native protocols: Fanuc, ABB, KUKA, Universal Robots
  • Per-part X / Y / θ to robot in <50 ms
  • Fixture-free picking for parts the camera can identify
  • Mixed-bin support: identify and pick the next valid part
Greenfield · spec from scratch

If you're designing a new robot cell.

Skip the fixture cost entirely. Spec a 2D vision station for flat-presented parts; spec a 3D structured-light station for bin-picking. The robot's pick path is computed from the camera data — no part presentation engineering required.

  • 2D for flat parts; 3D structured-light for bin-picking
  • Native robot protocols + vendor-agnostic ROS interface
  • Pick-path planning included — collision-aware in cluttered bins
  • Cycle time: 0.4-1.2 sec per pick depending on application
  • No fixture engineering required
How it works

Four steps from camera to verdict.

01

Image

Camera (2D or 3D) captures the parts in the workspace.

02

Identify

AI identifies each part: type, orientation, pick-feasibility, occlusion.

03

Plan

Best pick selected; pick path computed; collision check against environment.

04

Execute

Robot pose + pick command sent to controller. Robot picks; verifies; places.

Compatible with

The InspXion stack for this use case.

Ready to put this into production?

Get a sized configuration in 48 hours.

Tell us your line, your existing equipment, and your tolerance window. We'll come back with a retrofit assessment, a greenfield specification, and a rough cost estimate within two business days.