Increasing the degree of automation in handling with new technologies and thus raising leverage economic and qualitative potential: For this goal, we are developing handling technologies in the Deep Picking project to make depalletizing and picking tasks as robust and flexible as possible.
In intralogistics as well as in production processes, the handling of workpieces and products is an important but equally difficult task. There are many reasons for this, because the variety of products and the variety of workpieces is usually too extensive to set up the systems specifically for each product and to be able to guarantee a robust process.
In the Deep Picking project, technologies are being developed that are required both for use in intralogistic depalletizing processes and in kitting applications in production systems.
In the area of depalletizing, product bundles that are delivered on pallets are recognized by object position estimation algorithms and handled by roll-on gripping systems. Machine learning methods are used for object position detection in order to be able to robustly pick previously unknown objects. In addition, the roll-on gripping system from Premium Robotics GmbH is being further developed.
The existing bin picking software bp3TM is being further developed for the application of kitting. The focus here is on accelerating the process, increasing the robustness and the model-free gripping of unknown objects. This application is integrated in a robot cell and tested by an end user. These further developments with the help of machine learning enable the robust gripping of unknown, flat, entangled and complex objects.
Fraunhofer Institute for Manufacturing Engineering and Automation IPA