Using AI to reduce waste in the construction industry

On a construction site, building materials might arrive from anywhere in the world. “The construction sector is one of the most fragmented ever,” says Catherine De Wolf, SM ’14, PhD ’17, assistant professor of circular engineering for architecture at the Swiss Federal Institute of Technology Zurich (ETH Zurich). 

While that fragmentation might make the industry more productive, producing and transporting materials causes an incredible amount of carbon emissions. As buildings are demolished, they produce nearly 700 million tons of waste each year. 

“About half the materials we consume go into our buildings,” De Wolf says—and they account for about a third of all waste worldwide. So she’s employing artificial intelligence and other technologies to match builders with materials they can reuse from structures being demolished. 

When De Wolf came to MIT from her native Belgium in 2012 to pursue a master’s and PhD in building technology, global discussions on sustainable design focused on factors such as insulation and heating efficiency, not on the carbon emissions associated with the production and transport of a structure’s materials. 

“Everybody was talking about operational energy; nobody knew what ‘embodied carbon’ even was,” she says . But she collected industry data for the first benchmarking study on embodied carbon in buildings in 2014. Though many in the construction industry resist change, she was inspired by the “hacking culture” of MIT, where “you are taught that everything is possible, so when people say ‘Nothing is going to change,’ you say, ‘Let’s try it anyway.’”

Her lab at ETH Zurich has catalogued buildings around the world. It uses computer vision and machine learning to analyze buildings through Google Street View images, archival records, public demolition information, and auditing data. This enables the team to determine what materials within the buildings might be reused.

Her team has also deployed robots and drones to rapidly inventory materials at building sites before demolition. And they use AI to help architects draw up building plans based on materials available for reuse. 

De Wolf’s lab is now developing tools to rapidly match companies with other nearby sources of reusable materials for new projects and aims to create a larger platform to make similar connections across the world. 

“To really solve the problem of fragmentation in the industry, we must make the cataloguing, reuse, and assembly of materials faster and easier than the wrecking ball,” she says.