The smart robot crafting approach to computing materials

This study presents a universal method that combines robotic/mechanical automation with image processing and artificial intelligence (AI) to generate material models without any pre-existing knowledge of the material itself. Inspired by the “hand-eye-mind” process, used typically in designing and crafting, this study proposed a digital version of the process that is capable of automatically conducting a large number of material experiments, observing them using image recognition, and subsequently training AI. The proposed method generates neural network models for common digital design environments that help to bridge a wide range of design intentions, fabrication controls, and dynamic material behaviors. In this study, two different experiments were conducted using the same method. The first one generated a material model for the bending behavior of non-linear synthetic rubber, and the other involved the dynamic control of the form-finding process of thermoplastics based on dynamic annealing, which contributed to a new 3D printing method. With current progress, we are able to prove that such a workflow is a widely adaptable method that encompasses a large variety of material properties and fabrication methods. It enables design and construction using complex material behaviors without the support of existing material/structure models.