Home/Use case
Vision AI Cell
Use case · shown live at all about automation Zurich 2026 – two days of continuous operation at the booth.
The use case
Picking without programming
Conventional robotics needs fixed fixtures and pre-programmed positions – every new part means retooling and teaching. This cell shows another way.
A
The task
Recognize and handle arbitrary, changing objects – without fixed workpiece fixtures, without pre-programmed positions, without retooling between parts.
B
The solution
A ceiling camera captures the workspace, Vision AI determines coordinates, shape, geometry and orientation in real time and passes them straight to the robot. The robot picks camera-guided – and the hand-over sits ergonomically at grip height.
The cell in detail
Six stations, one system
From camera image to ergonomic output: click through the stations of the cell and see how the parts work together.
Tap the dots in the graphic
1Ceiling camera
Mounted above the workspace and calibrated to the cell’s coordinate system. One camera image replaces light barriers, hard stops and fixed workpiece fixtures.
2Vision AI
Determines exact coordinates, position and orientation for arbitrary objects – plus shape and geometry. The results are output in real time: to the robot, the controller or the control station.
3Robot
Picks, camera-guided, what Vision AI has measured – without pre-programmed positions. New parts need no retooling, just new seeing.
4Arbitrary objects
No fixed place, no fixed shape: parts may lie anywhere on the surface. What is recognized is the object itself – not the position where it is expected.
5Output station
The hand-over sits at hip height – ergonomically correct, with no reaching into the cell.
6Control station with MachineCopilot
Outside at the booth: ask the cell questions by voice, trigger a fault, watch the diagnosis live. The AI runs locally in the cell – no cloud.
Seen live
Two days of continuous operation at aaa Zurich
At booth G38 the cell ran non-stop – and in the Talk Lounge, the talk “The Autonomous Machine Assistant” presented the concrete fields of application on both days.
Result
What the use case shows
No teaching
New parts need no robot program – Vision AI recognizes them and delivers the grip data.
Position-independent
Objects may lie anywhere on the surface. What is recognized is the part itself, not its expected position.
Real-time output
Coordinates, shape, geometry and orientation are available instantly – for the robot, the controller or the control station.
Everything local
The entire AI runs on edge hardware in the cell – without a cloud connection, with full data sovereignty.
MachineCopilot
Your own pilot in 4–8 weeks
The standardized MachineCopilot pilot brings the AI assistant to a machine family of your choice – with a clear framework instead of open-ended project risk:
- 01Local edge deployment on your machine
- 02Defined documents and a clearly delimited use case
- 03Fixed scope, fixed price and agreed success metrics
Next step
In a no-obligation pilot consultation we work out where Physical AI gives you the greatest leverage in your production – concrete, honest and on equal terms.
Fixed scope · Fixed price
Contact
Let us talk about your machine.
In a no-obligation pilot consultation we work out where Physical AI gives you the greatest leverage in your production – concrete, honest and on equal terms.
Or reach us directly: +49 7805 9589-0·info@techtory.de
We usually get back to you within one working day.