From Sound to Structure: Machine Learning and Human Participation in PAFF
Temporary pavilion by Obuchi Laboratory, University of Tokyo. A “research by design” project combining machine learning, auditory perception, and participatory fabrication to create a cloud-like architectural form. Contributed as part of the structural team in close collaboration with Jun Sato Laboratory.
2017
Research by Design
PAFF – Projectile Acoustic Fiber Forest
Pavilion Design & Construction – University of Tokyo, 2017
Concept
PAFF (Projectile Acoustic Fiber Forest) was developed as a research by design project at Obuchi Laboratory, University of Tokyo. The pavilion investigated how auditory perception, machine learning, and participatory fabrication could be translated into architectural form. Digital fabrication is often associated with precision and automation; this project reintroduced human subjectivity and imprecision as design drivers, transforming data and perception into built structure.
Process
Participants were placed in a virtual 3D environment (Unity) and asked to locate sound sources by hearing. Their auditory data was recorded and processed through machine learning to identify patterns and differences in perception. This data was mapped to spatial coordinates and used to guide a participatory construction process: material (coconut fiber) was deposited by shooting it onto a mesh canopy, allowing the pavilion to grow as an accumulation of perception and action.
Design
The result was a cloud-like canopy supported by slender columns, opening views toward a ginkgo tree on site. The mesh caught fibers, forming a light, floating roof that combined digital data with the imperfect tactility of human participation.