COLLECTIVE DESIGNING FOR COLLECTIVE LIVING
Harvard Laboratory for Design Technologies | 2020
Research Project with City of Boston
Supervisors: Elizabeth Christoforetti
Duration: 2 months
Published (to be) on the Machine Learning and the City Reader for Wiley
Tools Used: Grasshopper, Rhino, HTML, CSS, Javascript, Pix2pix (ML model)
︎ Problem: With changing residential needs in communities in Boston, residents are forced to relocate from single family homes to triple deckers. Those buildings are not equipped to house multiple families and need to adapt to modern needs, while preserving the neighborhood identity.
︎ Solution: Through a web interface and a machine learning model, we are imagining co-design tools that will be used by some communities in Boston, and will allow them to customize and rethink their residential space.
*credits to Stanislas Chaillou for base website
︎ Problem: With changing residential needs in communities in Boston, residents are forced to relocate from single family homes to triple deckers. Those buildings are not equipped to house multiple families and need to adapt to modern needs, while preserving the neighborhood identity.
︎ Solution: Through a web interface and a machine learning model, we are imagining co-design tools that will be used by some communities in Boston, and will allow them to customize and rethink their residential space.
*credits to Stanislas Chaillou for base website
Problem
How might we design a tool that empowers residents to rethink their homes?
Machine Learning Model (pix2pix)
Basing the dataset on triple deckers, the goal was to allow manipulation of the typology while preserving the identity.
User generated buildings on Pix2pix Model
Questions
Multiple web trials were developed with the goal to mazimize user engagement.
How can we interface relevant user inputs from the community with the ML model?
To what degree does the community want to engage?
What are the criteria to be offered as design input?
What is the back-end vs front-end of the design tool?
Trials
1. Feature based
2. Component based
3. Action based
4. Tangible option
5. Hand drawn based
A tangible planning option
Try it out here: https://romysayah.github.io/UrbanStackMattapan/
Current Prototype
The Smart Planning Table is a design and communication that makes planning dense programs easier and more tangible. It translates physical notes into an architectural plan.
Node based coding with Grasshopper (Rhino), Python
Prototype
(in progress)
Prototype design mockup