New Research Shows How 2D Layouts Can Be Popped Up into 3D Structures
Imagine designing an inexpensive, flat and strong structure that bends perfectly into Spaceship Earth from Disney's EPCOT theme park.
NJIT's Przemyslaw Musialski, associate professor of computer science, believes his latest research is a step in the right direction, sitting at the intersection of architecture, computing and geometry.
Musialski calls this Generalized Deployable Elastic Geodesic Grids in a paper he presented with co-author Stefan Pillwein, at Vienna University of Technology, at the SIGGRAPH Asia 2021 conference last month in Tokyo.
"I am passionate about geometry and I'm exploring it at various levels. I'm also specializing in a field called computational fabrication," Musialski explained. Elastic refers to the structure's ability to expand like an igloo tent, where two poles are bent across and fixed, while geodesic means the expanded lines create the shortest path between two points on a surface.
Nature enables such techniques in everything from bamboo poles used to create huts to how a blade of grass bends when it's wet. But now, "The world of digital fabrication is growing. The hardware and the machines are getting better and better, now on the side of software we are going to keep pace with that," he said.
The research, sponsored by Vienna Science and Technology Fund, showed that it's possible to algorithmically make a quick prototype of a 2D structure and turn it into a deliberately designed, load-bearing 3D structure. Musialski is seeking new funding from the National Science Foundation to expand the research with twisting shapes and machine learning. His ultimate goal is to solve the computational challenges in real-time, and he imagines a future where this software is a standard feature in computer-aided design systems or a common real-life construction technique.
Musialski said that in his future research, he aims to extend the design process of elastic free-form structures with modern machine learning technology. It allows addressing the problem in two ways. First, novel machine learning techniques can solve specific physical problems, like elastic bending, faster and more efficiently. Second, generative neural networks can help the architect to explore the space of possible designs and suggest shapes based on already existing data.
Andrzej Zarzycki, an associate professor specializing in the intersection of architecture and technology, said he sees a bright future in Musialski's work. Zarzycki teaches in Hillier College of Architecture and Design and said he hopes to include such research in one of his courses.
"There is a natural synergy between what Przem and his Ph.D. students are working on, and architecture and design in general. In architecture, the concepts behind elastic geodesic grids could be utilized as adaptive and deployable lightweight structures," Zarzycki said. A notable building of this style is the Centre Pompidou-Metz, a modern art museum in Lorraine, France, he noted.
"They provide significant advantages for architects and builders as these structures could be mass-produced, yet customizable, easily shipped and rapidly deployed as emergency shelters during major disasters. These concepts could also be used on a much smaller scale as adaptive building facades with kinetically changing fenestrations and textural characteristics," Zarzycki continued. "The process and computational tools Przem and his Ph.D. students developed bridges creative form-making with constructibility and feasibility evaluation."
