Coordinated movements are a complex field. This is already reflected in the fact that we humans need years until we have the necessary practice. Researchers at ETH Zurich have now succeeded in optimizing robots so that they can guide a tool precisely. Simon Düser belongs to the research group of Stelian Coros from the Institute for Intelligent Interactive Systems at ETH Zurich. Together with colleagues, the scientist has developed a hot-wire cutting robot "RoboCut". The wire can be flexibly bent as you work, making it possible to produce much more complex shapes with far fewer cuts than was possible with previous systems. Conventional cutting robots have a stiff, electrically heatable wire. It was only possible to process so-called ruled surfaces that contain a straight line at every point.
Demanding optimization tasks that are integrated into an algorithm are required to predict an optimal movement of two robot arms that precisely guide a tool. For Simon Dünser, that was exactly the focus of the project: “The special thing about RoboCut are the complex optimization calculations. These are necessary to find the most efficient toolpaths possible and at the same time to melt the desired shape out of the Styrofoam block as precisely as possible.”
Shaped in ten cuts sitting bunnies out of styrofoam block.
The developed RoboCut can create indentations in a plastic block. It is not limited to planes, cylinders, cones, or saddle surfaces. This makes it flexible to use. According to the researchers, its greatest advantage lies in the fact that the targeted bending of the wire requires far fewer cuts compared to the method with ruled surfaces. For example, the researchers have succeeded in forming the figure of a sitting rabbit out of a block of polystyrene using the bendable wire in just ten cuts. The outline of the rabbit was clearly recognizable after just two cuts, and the result after ten cuts was comparable to a woodcut.
The aim of the researchers is not only to fundamentally improve the traditional hot-wire method with the RoboCut project, but also to develop concrete application goals. From her point of view, a use in architecture is conceivable, in order to produce individual molds for concrete parts from Styrofoam. This would make it possible to design facades with more variety or to develop new building block systems.
With three optimizations to the goal.
For the series of tests, the scientists attached the wire to a two-armed ABB Yumi robot. In this way, they wanted to ensure that the wire could be moved in a controlled manner. This was followed by a calculation of how the wire reacts to the movements of the robot arms. With the help of simulations, the researchers determined the positions that lead to unstable wire positions or where there is a risk of wire breakage and ruled them out. These simulations served as the basis for the actual optimization. To do this, three interrelated aspects had to be taken into account at the same time. The first was at the physical level: researchers had to predict the controlled bending and movement of the wire to make the desired cuts. The second aspect focused on the shape: The aim here was to determine a cutting sequence with which the surface largely resembles the target shape in as few cuts as possible. Thirdly, the robot also had to be considered, so that collisions with parts or the environment can be avoided and there are no accidental cuts.
Simon Dünser succeeded for the first time in his RoboCut project in finding out these various parameters of the complex task and integrating them at the right point in a global optimization algorithm. His way: a structured method that he designed specifically for this purpose. They are based on an overriding objective: the wire should always cut as close as possible to the surface of the target object. He then assigned any constraints to costs and optimized them as a sum. Since such calculations always involve local minima that influence the final result, Dünser approached the desired result in several individual steps.
Clever plastic cutter creates tomorrow's concrete construction technology.
The scientist is certain that his methodology is also suitable for other areas of application, for example for planning tool paths in other cutting and milling techniques. It is particularly suitable for complex, rotationally symmetrical shapes. Then his method opens up a much larger scope for the simulations. It could also bring progress in spark erosion, because such complicated and therefore more efficient cuts are possible than with today own stiff wires. Together with the EPF Lausanne research group, Dünser is planning a specific application: They want to use a large version of the hot-wire cutting robot to develop systematic building blocks for building structures that do not require mortar or fastening technology. It is important that these elements themselves hold together stably. The robot should then also cut Styrofoam molds with which the various building blocks are cast in concrete.