Easy-to-use, the New Oil and Bin Picking – About Current Challenges in Automation

For us it is easy to grab unsorted items out of a bin – be it handkerchiefs, candy or screws of different sizes. But it’s not for a robot. “Bin picking” is at the top of the agenda in the automation industry. It is also extremely interesting for the development of a robust algorithm for the discipline of machine learning – and therefore one of the topics in the KUKA press conference at Automatica 2018 in Munich.

They meet in the halls of Messe München: experts from the industry and interested journalists. They are swapping out through app stores for the industry and shared services, through Big Data & Smart data. They talk about global concepts for the use of artificial intelligence and the different regulations of the individual countries. They discuss the sensible sorting of the data, the access to it and bin-picking with sensitive robots.

“Cobots are definitely in the spotlight of this year’s automatica,” begins Christian Tarragona, Senior Vice President of KUKA’s research and development, his introductory statement. Next to him are with David Fuller, Chief Technology Officer at KUKA, and Dr. Christian Baur, CEO of Swisslog Logistics Automation two other experts. As a trio they are facing the questions of the journalists at the KUKA press talk.

Data is the New Oil
“Those who shut themselves off to software today have already lost,” says Christian Tarragona, thinking in particular of the new LBR iisy, which as a sensitive lightweight robot which combines the topics of human-robot collaboration, sensitivity and easy-to-use. “The focus must be on the fact that the robot in combination with software is easy to use, secure, connected and smart,” adds David Fuller. “This applies not only to the area of Cobots, but to all industrial robots.”

Smart builds the bridge to artificial intelligence, to analyze the right data. “Data is the new oil. The internet is full of standardized file formats for images – like .jpeg or .png which was great for learning image-based object recognition. However, there is no standardized file format for bin-picking or spot welding and other key processes. For the application of artificial intelligence you need all of this kind of data. We are focused on collecting this kind of process data and optimizing performance via applied artificial intelligence,” says David Fuller.

What robotic systems have to learn from the human hand
In warehouses filled with everyday consumer goods, driverless transport platforms drive entire shelves to output stations. There are robots waiting to pick the goods out of the bins and to assemble them according to the order. “In the field of artificial intelligence it is necessary to make things parallel. In general, it is necessary to consider which data is needed and how it can then be analyzed. Because an infinite amount of data does not help if it is not the right data,” says Dr. Christian Baur.   ”As soon as I have the right data, I can solve specific topics like for example bin-picking.”

What makes bin-picking so complex? It is the interaction of hardware with the right software. In addition to a sensitive robot, cameras and a matching gripper are needed, which can accommodate handkerchiefs, candy and screws. Our human hand is clearly in the advantage of this task. So it must be the goal that the whole system learns – not just only the robot. Only then the system can effectively play together and learn, and then at some point know that it has to deal with a glass bottle more cautiously than with a plastic bottle.

Related Posts