Picking across a changing product mix
E-commerce fulfillment brings a wide range of product sizes, shapes and SKUs into the same picking process. Glossy, thin and semi-transparent objects add further difficulty for vision and gripping.
Eureka’s masterless 3D processing identifies objects without CAD models, registered reference images or training for each item. Its AI-based stereo camera works without pattern projection, reducing sensitivity to reflections and lighting when recognizing these objects.
Choosing the grasp and planning the motion
The system determines an object’s position and orientation, then selects a suitable grasp. Bridgestone Soft Robotics’ TETOTE and gripper adapts by using either suction or four fingers, depending on the item’s characteristics.
AI-generated grasp poses feed into robot motion planning, with collision avoidance and paths generated in real time. Integration between 3D vision and robot control simplifies precise calibration and supports flexibility in the choice of robot and end effector.
What the collaboration demonstrated
The collaboration demonstrated the feasibility of automated piece picking across varied SKUs, including adaptive selection between suction and fingers. Reported outcomes include fast, reliable handling across a wide product mix, supporting the approach’s potential for scaling piece-picking automation in e-commerce fulfillment.
