Automation is constantly advancing. Robots are taking on increasingly diverse tasks in factories and other settings. They have to be taught each of these individual tasks first, though, and that’s no trivial matter. It takes specialists versed in complex robotics programming. The major robot manufacturers offer a wide range of training courses. According to the International Federation of Robotics, ABB, FANUC and KUKA train between 10,000 and 30,000 participants in more than 30 countries every year. But knowing the programming language of a single manufacturer is often not enough. Robots from multiple suppliers often stand side by side in factory halls – and all of them work with their own specific code. All this makes programming and commissioning a robot costly and time-consuming. In fact, these aspects account for over 30 percent of the planned cost of a robot system.
Learning via imitation instead of programming
The need for ideas on ways to simplify programming is clearly huge. A few companies are already giving it a go, and they mainly come from Germany. Wandelbots is one of a number of startups that want to make it easier for companies to use robots. “The founders of Wandelbots come from the IT sector and were shocked 10 years ago when they saw that operating concepts from the 80s were still being used for industrial robots,” says Jan Drechsler, head of marketing at Wandelbots. “They knew there had to be a better way.”
The Wandelbots approach to this problem is “no code”, meaning that if you want to teach a robot a new task, you don’t have to program it anymore. Instead, you can use a sensor-equipped pen – which Wandelbots calls a “TracePen” – to demonstrate the motion that the robot is supposed to execute. For a welding robot, for instance, a user would simply trace the line they wanted to have welded on a workpiece. In the corresponding app, they can set the specific parameters and make fine adjustments to the path drawn. The software then generates the code in the robot’s respective programming language. And it’s not limited to just one language, either. With the Wandelbots system, users can teach robots from all the major manufacturers without writing a single line of code.
According to Drechsler, a robot can be programmed 70 times faster with the help of Wandelbots than in the conventional way. This was one of the findings of a pilot project at BMW, where the task of gluing windshields was taught to a robot using the TracePen.
Drechsler also reports that costs can be reduced by 90 percent thanks to the Wandelbots software. This is because the company doesn’t need a system integrator that often not only installs industrial robots, but handles their programming as well. Since the programming doesn’t take as long, robot downtime and the associated halts in production are also reduced.
Software translates between human and robot
All these advantages notwithstanding, Drechsler believes the software’s main strength is its ease of use. “Anyone can teach a robot with it,” he says, adding that this even includes production employees with no robotics experience. “Instead of just having just one expert in this field, a company might now have 20 employees working with robots at the same time.” The decisive factor lies in how the Wandelbots software does the “translation” work between people and machines. In theory, it could also be used with other input devices. For example, the founders of Wandelbots initially started working on a smart jacket that an operator could wear to teach a robot movement sequences. When the user lifted an arm, the robot made the same movement.
However, the use of intelligent clothing for programming was too complex at the time, as Drechsler explains. “You need different sizes and the jacket eventually needs to be washed. A pen, on the other hand, is much closer to how we actually handle tools.”
Intuitive operation thanks to artificial intelligence
The startup drag and bot also wants to simplify working with robots. A spin-off of the Fraunhofer Institute for Manufacturing Engineering and Automation IPA, it offers companies a package of ready-made software components that can be easily assembled into the desired robot program via drag and drop without any special robotics expertise. Its drag&bot software can also be used for robots from different manufacturers.
The same goes for ArtiMinds. Like drag and bot, this startup is a spin-off of a research institution – the Karlsruhe Institute of Technology (KIT). ArtiMinds takes a similar approach, as well. Users can create a complex robotics program from pre-configured components without having to write code themselves. Those familiar with it report that the software is so intuitive that no training is required.
Intuitive operation is also the order of the day at Micropsi, a Berlin startup that uses machine learning to train robots for their jobs. Here, the user shows a robot its task by guiding its arm by hand. These movements are recorded by a camera and a force torque sensor, and then processed using artificial intelligence. In this way, the robot learns more than just the activity in question. It’s also able to react independently if the situation changes – if a part it needs is in a different location, for instance.
Robots can learn a variety of different tasks with the systems from ArtiMinds, drag and bot or Micropsi. This can include various assembly activities or pick-and-place tasks – that is, picking up parts and putting them down again at another location. Wandelbots, on the other hand, is still concentrating on tasks where a robot has to follow a path, such as welding or gluing. “In welding, there are many curves or long paths where accuracy is crucial,” explains Drechsler. “We can map this quite well with the TracePen.” He points to the lack of skilled workers as one reason why there is so much pressure to automate in this area.