Together with its new software image 25.01.0, Roboception has also launched its software product rc_reason ItemPickAI.

The new rc_reason ItemPickAI software enables an AI-based grasping a precisely oriented placement of so-called ‘unseen’ objects – i.e. objects, that the system has not explicitly learned by teaching models or previously seen in training data.

ItemPickAI enables an automated separation and an oriented placement even of deformable objects: Robotic palletizing and unpacking of bags, sacks and various pouch packaging can hence be implemented precisely and efficiently.

This new AI-supported solution for robotic pick-and-place applications using suction grippers calculates grasp points on unknown, deformable objects. These objects can be coming from chaotic boxes or load carriers; unmixed and mixed configurations are possible, as is an oriented placement.

A vision system comprising a 3D stereo sensor (rc_visard/ rc_visard NG or rc_viscore) and the ItemPickAI Module reliably detects and segments unknown objects with uneven surfaces and deformable filling. It provides a robot with precise grasp points for an efficient handling of the objects via a suction gripper. A robust segmentation of the units is ensured by the use of neural networks.

Applications lie not only in classical automation, but also in the field of logistics: Using ItemPickAI, a robot can pick disordered and crushable objects from a crate and place them in an aligned position, or unpack pallets and place objects on e.g. a cross-belt sorter.