Team: Theo Dattola, Emil Senheusen, Yibo Qiao, Isabella Socorro
Project: Design Research
Advisor: Casey Rehm
Artificial Inteligence Produced Proto-Architecture
The project utilizes convolutional neural networks in combination with robotic assembly algorithms to produce scaled timber structures. In addition to 3D form generation. Neural networks will be used in to develop 2D texture applications for the structures, exploring the relationship between 3D form and 2D image. The class will explore the opportunities and limitations inherent in machine learning-based automated design and fabrication.
Additionally, the class will explore how mereological explorations within architectural production can be utilized as generative inputs into non-human forms of graphic expression.
The semester long project has three main components. The first exercise will focus on training 3D GAN convolutional neural networks on student created datasets of obj files to produce massing models of proto-architectural objects. These models will then be translated into pattern schedules for wood tectonic elements and assembled utilizing the robots.
Pattern and tool path generation will utilize Grasshopper and Python with initial techniques utilizing cellular growth algorithms. Students will explore how to subdivide their massings into rational sub-components to accentuate pattern in relationship to form and to optimize orientation for the robot work sphere.
The second phase of the project will utilize photographs of the assembled components as inputs in Cycle Consistent Adversarial Networks to generate a graphics package for the final objects. These will be applied through vinyl printing, flatbed printing, hydro dipping or robotic airbrushing. Students will exhibit their final object of approximately 1m3 as well as process samples of pattern study and neural network output.
In the end, the project utilizes several different mechanisms to produce a proto architecture made out of discrete components, produced using neural Gan Networks known as Artificial Intelligence, and the Robotic Arms to construct the output for said neural networks. The role of the designer is highly questioned in this exercise and puts us in the curator role more often than not.