Famous models are the hanging gardens of Semiprimes, known from antiquity. The replica, a 22.5 meter high planted sculpture done by robots. ETH Zurich’s research group uses artificial intelligence and four interlinked robots.
The task that researchers from the group of ETH architecture professors Fabio Gramazio and Matthias Kohler have set themselves together with the companies Müller Ilene Landscape Architects, Timbale, and other partners from industry and research is not easy: five geometrically complex wooden bowls, a total of 22.5 meters high, which are arranged in a slightly networked manner and supported by eight slender steel supports. They have chosen the hanging gardens of the Semiprimes, known from antiquity, as a model.
The architectural sculpture, made with innovative, digital methods, will be used in the Tech Cluster Zug. Behind this is the idea of creating an innovation quarter on the premises of the company V-Zug AG in the city of Zug in Switzerland. Metals Zug AG sponsors the project.
Hanging Gardens: Diverse Design Options With AI And Four Robots
If the aim is to consider the different requirements for a building or a structure, architects adopt them as part of a classic design process until they are optimally fulfilled. In the Semiprimes project, a conscious decision was made to take a completely different approach: A tailor-made machine learning algorithm was used – developed in cooperation with the Swiss Data Science Center. The advantage is that it opens up a wide range of design options for researchers. “The computer model enables us to reverse the conventional design process and to explore the entire scope for designing a project. This creates new, often surprising geometries,” says Matthias Kohler, Professor of Architecture and Digital Fabrication at ETH Zurich.
AI: On the Hönggerberg campus, the researchers explored the designs together in three dimensions in the so-called “Immersive Design Lab,” a laboratory for augmented reality, and worked on them in real time. In addition, they used software that made it relatively easy for them to adapt the designs of the wooden bowls. As soon as the researchers changed a single point within the geometry of a shell, the software automatically adapted the entire geometry to this change. And the bowls consist of around 70 wooden panels. At the same time, the software can also consider the relevant production parameters. This includes, for example, the maximum possible weight of a plate. In this way, it always creates the most efficient and resilient configuration. The Computational Robotics Lab developed this software at ETH.
AI Makes Robotic Arms Dance
After trying out different variants, the researchers finally came up with the optimal design. This is now to be implemented in the robotic production laboratory at ETH Zurich. If you look into this laboratory, you might think that four robotic arms are dancing with each other in a precise manner. They move with the same rhythm, accept the wooden panel that has been assigned to them, and finally place it in the room as the computer design dictates. An algorithm calculates the precise movements so that the robot arms do not collide. As soon as the machines have placed four panels next to each other, artisans temporarily connect them. Then they are glued with a special casting resin. The robot arms assemble between 51 and 88 wooden panels into a finished rigid shell.
Robotic production has several advantages – especially compared to conventional construction with wood. The robot lifts heavy parts here. He is also able to position the components much more precisely. Particular substructures are customarily required for the assembly process. This can be dispensed with, minimizing the effort and using fewer resources. The project in Zug should be finished by spring 2022. Therefore, the individual bowls are currently being brought to Zug by truck. After completion, the planting will follow so that the hanging gardens can inspire visitors in summer 2022.
ALSO RAED: Start The New Year With Certainty: 5 Best Tips For IT Admins
