The challenge
Panel radiators of different types and sizes had to be picked safely off a roller conveyor, every product had to go to the right pallet, and the work had to be balanced across three robots.
Our solution
A camera finds the radiator the robot will pick: a deep-learning semantic segmentation model separates the radiator from the background at pixel level, even on the glossy, striped surface of the roller conveyor.
Three ABB robots positioned along the line grip the panels with grippers we designed and built for this product and stack them on pallets. The grippers carry the Pi Robotik name.
The line runs on a management interface we developed: each radiator's type and dimensions are identified from its barcode, the product is assigned to one of the three robots, and the robot stacks it on the matching pallet. The workload split between the robots is set on screen; pallet fill levels and production counts per robot and per panel type are monitored live.
