Skip to main content

SCI-Arc Faculty Awarded Google Grant for AI Housing Research

In October, SCI-Arc faculty Mimi Zeiger (M.Arch '98) and M. Casey Rehm were announced as recipients of Google’s $20,000 Artists and Machine Intelligence Research Award to fund a project focused on ways to employ AI in an effort to better understand the possibilities of Accessory Dwelling Units (ADUs) to mitigate the affordable housing crisis in Los Angeles.

aerial view green blocks
Infrastructure Prosthetics. Located on brownfield sites along the Alameda Corridor, these platforms attach to existing city infrastructure to allow for gradual, affordable development over time. The site strategy is determined by using AI to find the shortest path on the site based on the existing grid lines overlay from the city. The generative system helps the developers and designers in developing a platform that densifies the area with the natural growth of housing units. Student team: Julia Pike, Lourenco Vaz Pinto, Philippe Maman, Yash Mehta. Instructors: Mimi Zeiger and M. Casey Rehm. Spring term 2020, SCI-Arc.

The project, entitled Backyard Home Data Explorer, is part of an ongoing research initiative of SCI-Arc’s Urban Pasts and Futures Lab that looks at how machine learning and Neural Network (NN) platforms might be leveraged to address critical issues within the built environment, with specific interest in the future of equitable housing. According to the proposal presenting the initiative, the project seeks to engage how “research and platform development pairs machine learning with housing policy and design with the intent of finding creative ways to reinvent and disrupt bureaucratic systems that often fail to keep pace with housing needs and the changing culture of cities.”

There is growing local interest in ADU dwellings due to recently passed bills (AB 2299, co-authored by Dana Cuff, an architect and director of cityLAB at UCLA in 2016) as a means to create more abundant affordable housing. SCI-Arc’s project will analyze the feasibility of introducing ADUs on specific sites and prototype an interactive app that allows the user to simulate variable impacts, from policy-based to environmental, on the city. While the work has been focused on LA, the app could be applicable to any city.

blue tiles computer render black background
House.LA proposes a future housing solution in the form of an app-based matching service, where a digital platform pairs owners and renters of ADUs. AI technologies, such as neural networks and video game design applications, are deployed throughout all stages of the design process and matching service: from sourcing lots where there is preexisting space for an ADU, to placing utility cores effectively while avoiding existing trees and foliage, to the tectonic arrangement of component parts around the core in configurations appropriate to site and user, as well as in the digital app as a way to match owners and users with particular quantitative and qualitative needs. Student Team: Luke Falcone, Zihan Gao, Sadvi Jayanth, Amanda Kotch. Instructors: Mimi Zeiger and M. Casey Rehm. Spring term 2020, SCI-Arc.

The Google-funded project is a continuation of work Zeiger and Rehm began in spring 2020, when they co-taught a seminar dedicated to changing housing developments in the city. Preliminary work was exhibited by Pando Populus in the summer of 2020, where it was singled out as “environmental, technological, philosophical, and moral.”

The research activities associated with the project are being integrated into SCI-Arc’s curriculum. This semester, Rehm and Zeiger are teaching a seminar entitled AI+ADU: Approaches in Housing and Neural Networks. The class will also work to identify policies at county, state, and federal levels that can be optimized for use with AI and machine learning tools, which will also look at historical bias within housing development and in AI.

Zeiger explains, “This project reflects a growing emphasis at SCI-Arc on engaging students in real-world, contemporary research that relates to some of the critical issues of our time.”

circle architecture render grey yellow white
Students Jules Benillouche, Jennifer Rufener, Andrew Stone, and Leo Wan integrated neural networks using trained typical ADU footprints with slope data to design housing solutions on difficult terrain.

Starting this August, research assistants together with a web developer will build an open source, online platform for policymakers, designers, and the general public, which will use data visualization methods, machine learning, and AI to provide numerous variations of the size and distribution of ADUs based on relevant variables. Data will be collected from several sources including the City of Los Angeles GIS Hub and social media platforms.

In terms of the technical aspects of Backyard Home Data Explorer, Rehm, who is also coordinator of SCI-Arc’s Architectural Technologies postgraduate program, will supervise and advise the AI platform development. “We will be placing students from the Architectural Technologies program in research assistant positions, who will be assisting Mimi and I in the spring seminar and in building the research for the AI component,” notes Rehm.

A website has been created to track progress on the project and upcoming course at