Cobotic PoseStream

Cobotic PoseStream is a deep learning convolutional neural network Object Pose Estimator. Precise pose is retrieved for visible objects in real time. PoseStream locates objects even in challenging lighting conditions and cluttered or occluded scenes. PoseStream plans a suitable grip and approach path for picking the part. PoseStream provides a Python API that can be connected to a robot controller for bin picking or precise assembly of gripped parts

Speed: 16 FPS 

Accuracy: 1mm / 0.3mm (Cluttered / Singulated)

Use cases: Pick parts from bins, trays etc. Precision assembly or placement of parts to PCB, tray etc. Inspection of part placement

Cobotic SmartTweezer

Cobotic SmartTweezer is a servo controlled tweezer gripper for ABB YuMi robots. The camera is positioned so it can see parts in the tweezer gripper as well as search for parts in a bin. The tweezer servo is controlled as an additional axis by the robot, ensuring full synchronization and coordination between tweezer and robot motion. Our solution works with any tweezer, the exact tweezer used can be selected based on application needs*. The gripper has an absolute position potentiometer ensuring that position and calibration is maintained even when the gripper is powered off. Our cable dresspack provides the needed connections for the smart tweezer with minimal restrictions on the robot movement.

Grip Force:  10N

Grip stroke: 25 mm

Speed:  0.3s full open or close

Positioning accuracy:  0.1 mm

Weight: 300g

Robot compatibility: ABB YuMi**

*For the Automatica 2023 trade show we use Knipex 92 78 77 ESD Universal tweezer

**Upon request we can adapt the tweezer gripper for other ABB robots. If syncronized motion is not required, the tweezer gripper can be adapted to any robot brand.*92 78 77 ES

Cobotic Bin2Board


Bin2Board is our solution for seamless bin picking and assembly of through-hole assembly components. The robot picks randomly oriented parts from a bin, corrects the orientation with a handover and inserts the components to the PCB. The assembly tolerance for correct insertion is 0.3 mm.  The robot also inspects the assembly to ensure that the end result is correct. All steps of the assembly process are guided and controlled by Cobotic PoseStream.

Bin2Board can be used in electronics manufacturing as a flexible alternative for low to mid volume production where investment for conventional automation methods are not justified .

Bin2Board uses standard tweezers and the assembly is controlled by software rather than fixed automation. This means that the same system can handle multiple different articles, as long as PoseStream has been trained to recognize and locate the article.

Article changeover time: 24 h to train our neural network on a new article

Cycle time: *


*The video above and our Automatica 2023 trade show demo are a showcase of the possibilities unlocked by AI guided collaborative robots. We believe that a reasonable expectation for a production ready system would be around 25-30 parts/minute for low accuracy insertions (bin to tray, bin to lead cutter, etc) and 10-12 parts/minute for precision assembly to PCB.