Robots | Artificial Intelligence | machine learning and cameras are becoming a must

Kuka relies on more flexibility in automation: why the demand for the combination of robot and camera systems is currently increasing so rapidly.

Robotics and automation are finding their way into more and more areas in the industry, medium-sized companies and trade. At the same time, seeing robots, i.e. the use of camera technology on industrial robots, are experiencing increasing demand. Because cameras make robots flexible, that end customers the enormous variety of variants in their production.
“For us, it’s about robotics in medium-sized companies and flexible production – we face this challenge again and again and that’s why we also use sensors,” says Michael Suppa, Managing Director at Roboception. This topic falls under Industry 4.0 because there is a flexible product that has to be designed, in which the sensors play a major role and the processing of the corresponding information.
“We have the scenario of a large, constantly changing product portfolio and variants in production that have to be managed via the same system. Something like this is rather difficult with classic automation approaches. In addition, customers today are more and more interested in plug & produce solutions get,” explains Suppa.

Michael Suppa is the managing director and co-founder of the Munich company Roboception and an expert in innovative 3D perception solutions. As a pioneer in the field of 3D sensor technology, Roboception combines classic and AI-based methods so that robotic systems can capture their environment in real-time. They enable robots to see and make decisions, so to speak – central elements for future-oriented and flexible automation solutions

The robot’s camera technology recognizes 100 different variants

A practical example is provided by Kautenburger GmbH, which, in cooperation with Roboception and Kuka, is developing a solution for a Spanish customer to stack and depalletize a wide variety of kiln bricks on kiln cars and pallets. Software modules and sensors from Roboception and robots from Kuka are used, with stereo cameras being used, as this is no longer a problem in terms of computing power.
“There are up to 100 different types of oven bricks,” says managing director Christian Kautenburger, “and without camera technology, there are many false trips and collisions. With camera technology, there are no missed trips or collisions with a cycle time of nine seconds. The cycle time was previously 18 seconds. In addition, there were clearing times of half a shift in the event of any collisions.”
At first glance, the stack of stove bricks looks very orderly. “However, the transport route to the kiln and the shrinkage behaviour of the stones in the kiln makes the use of camera technology necessary, because the position of the stones for the robot cannot be defined precisely enough on its own,” says Kautenburger.

Christian Kautenburger is Managing Director of the Saarland Kautenburger GmbH . Together with Roboception, the medium-sized company has implemented an application for depalletizing furnace bricks. Software modules and sensors from Roboception were used for this. Further projects with applied AI solutions are being planned. 

Automation attractive even for the smallest companies

“The camera and robot system was very easy for us to commission. Of course, this simplicity also helps inexperienced users from medium-sized companies to think more about automation,” says Kautenburger. His customers are currently facing two challenges: cost pressure and a shortage of skilled workers.
” There is now also a shortage of skilled workers at the relocated locations abroad. Customers are considering bringing production back by automating it here. A customer in Romania actually has a complete line due to the shortage of skilled workers. The customer now wants us to do it as an automated solution. Here, too, we need a solution that sees,” emphasizes Kautenburger.
In addition, there is also the quality: It is not at all about saving personnel or costs but increasing the quality, which is too bad in manual work.
“I’ve noticed that even the smallest companies, for example in the machining or component deburring sectors, are thinking about using robots. 

Machine learning and 3D sensors make it easier to use

Nobody needs to be afraid of taking the step of using robotics. “Because machine learning and 3D sensor technology give us an important basis for realizing flexible production: we are replacing mechanical solutions or feeds with sensors and software, so to speak. Machine learning is a method for finding parameter sets from a large variation,” describes soup.
There is either a data-based or a model-based approach to how this parameter search is to be carried out. Suppa explains: “If I don’t have any model data, i.e. unknown objects, I have to resort to data-driven approaches, which means I record objects and train them from this data and then generate templates or reference models. In the context of industrial automation, where we usually always have CAD data, I can generate the reference models using a simulation. The advantage of the model-based approach is that the customer does not need any training time on site.”

Amortization of the costs for robotics and sensors

According to Kautenburger, the investment in automation usually has to pay for itself within a year. “That’s not possible here because it requires one robot for palletizing and one for depalletizing, which replaces six jobs and requires an amortization period of three to four years,” explains Kautenburger.
But in the refractory industry, there are very conservative investors, for whom an investment in the future is always valuable. In addition, the quality increased in terms of process reliability, since the robot works very carefully and the employee does not cause any damage to the stones such as broken edges.
“We have achieved a payback period of seven months for a customer who wants to improve quality,” Kautenburger calculates another example.

Keep the hurdle for the use of robotics low

Finally, Suppa reports which hurdles need to be cleared for potential users: “Due to the increasing number of variants, customers are willing to use sensors – on the premise that they don’t become too complex. Because the user doesn’t want to deal with the ‘The task of image processing. That’s the motto: Hide complexity and allow usability.”