Design Professor Embraces AI to Quantify Accessibility of Buildings
Beauty is in the eye of the beholder, but walkability could be evaluated by algorithms, according to new research from an unconventional professor in NJIT's Hillier College of Architecture and Design.
"Most people assume that architects have some tool to visualize or analyze how people will use the building. It's not true, they don't," explained Assistant Professor Mathew Schwartz.
Architects might anticipate walking paths, guess natural routes based on experience, or quiz the future occupants about their intentions, but the idea of using software to calculate the physiological metrics of accessibility in a room, building, or city has been an exotic and seemingly impossible quest for a long time. Metrics could include biomechanics, energy and time. It's even possible to calculate a score for nuanced concerns such as civil rights, safety and views, given the availability of data on exclusion, crime, or solar/lunar activity, Schwartz noted.
Researchers have considered that level of artificial intelligence for a long time, but it never seemed sharp enough — "It's cutting-edge, but it's been cutting-edge for 15 years," Schwartz said.
His new algorithm might fix that dilemma. Schwartz calls it SHAPE — Spatial Human Accessibility graph for Planning and Environment analysis. It uses the same digital models that a designer builds in their existing software, employs a computer graphics method called ray tracing to compute the connections between designated points in question, and outputs 3-D graphs used to calculate routes that humans may not have considered. The software could be a useful tool for positioning furniture, finding safer emergency exits, or developing more efficient paths through airports. His work has been supported by the Army, which presumably could use it to evaluate movements of friend and foe.
Drafting tables and T-squares were long ago replaced by software and mouses, so it's illogical that software is not yet widely applied to human accessibility by architects, Schwartz asserted. Health and safety training for architects is generally limited to air quality, emergency exit locations, wheelchair ramps and other factors that are determined by building codes. But the considerations of human factors should be the domain of architects and designers, he said. "If it's not the architect, then who? If it's not the one making and designing the building, then who is going to be the one sitting there saying it's not good for this person? There's a difference between saying people consider it and people quantifying it."
Schwartz noted some incremental progress toward such goals. Zillow, a popular real estate website, includes what they call a Walk Score for home listings — but it's focused on proximity to mass transit, parks and schools, not what's inside residences. Another example is researchers in France who built a mobile application for runners to find the best routes. They cited Schwartz's earlier work on similar topics.
Schwartz said he expects walkability quantification, based on machine learning and determined by age or health, to eventually become a standard tool in building design software. That could happen in 5-10 years from now.
"There's a lot of things I want people to know. If people know what's possible, in order to make a design better, they should demand it from cities and governments when public funds are being used for those spaces. If we spend billions on Newark Airport, that airport should have gone through some quantitative analysis and evaluation to ensure that people of all abilities are able to comfortably and efficiently navigate it," he observed.
On campus, "I teach human factors every semester. It's a required class for the art and design school, every student. I think a lot of the work I'm doing now will likely be integrated into what the students use in 6, 7 years."
The algorithms and methods are described in technical detail for Schwartz's latest published paper, "Human-centric accessibility graph for environment analysis" in the July 2021 issue of Automation in Construction. Though not listed as a co-author, he also worked with Assistant Professor Margarita Vinnikov, of NJIT's Ying Wu College of Computing. "She was working on augmented reality interfaces that would visualize the data calculated by my method," he said. "Her past work dealt with transportation research and human vision, but we are also continuing to work together on these topics."
