Digital twins simplify underground pipe inspection

There’s a lot we don’t know about the location of underground pipes—but advances in AI tools are making it easier to map, and perhaps more importantly, understand them, according to Howie Choset, creating a pathway for inspection and discovery of damage like corrosion and natural gas leaks, issues that can be catastrophic to human safety and detrimental to the climate when emissions seep into the atmosphere. As a type of underground infrastructure, pipes facilitate the transfer of natural gas as an energy source to homes and businesses—as well as handling fluids like water and sewage. 

“We want to create a 3D, Google Earth-like rendering that lets us see where all the pipes are,” said Choset, the Kavčić-Moura Professor of Computer Science at Carnegie Mellon University. “Our robots can create high resolution digital twins of underground pipes that uncover problems that crawler robots alone couldn’t—in fact, our pipe data is so great, we can geolocate the pipes that we’re mapping in the moment, as we’re mapping them.” 

To boost the efforts of his CMU spinoff company Pipe Force AI, co-founder Choset and his graduate student researchers in the Biorobotics Lab are using support from the Scott Institute for Energy Innovation entrepreneurship award to conduct pilot studies, attend pipe inspections, and interact with potential customers around the US.

Our robots can create high resolution digital twins of underground pipes that uncover problems that crawler robots alone couldn’t.

Howie Choset, Kavčić-Moura Professor of Computer Science, Carnegie Mellon University

The inspection companies that make contracts with utilities and the construction firms that select job sites are two stakeholders who would benefit from precise knowledge of pipe location and condition to protect workers and save money by assessing damage before work begins. 

Choset’s team is starting with data collection from storm drain pipes, which are quickly exceeding their capacity as weather events become more extreme and cause greater damage, as a first step in brokering partnerships with municipalities to assess their underground infrastructure. 

For graduate students whose work mostly happens in the lab, customer discovery and field studies are important exercises for meeting in the middle to understand how technical prowess must complement community needs to create an effective hardware or software solution. 

“Whenever you see something being sold to a customer, it seems like it’s common sense,” said Choset. “But connecting those two dots from what you have to what they want is really hard.” 

Choset notes that Pipe Force AI has switched its focus from building dexterous robots in favor of facilitating access to data that creates high-resolution digital twins. The technical work revolves around developing sensors that sit in a mapping box and attach to robots—the focal point of the company’s future in commercialization.