Researchers at Carnegie Mellon University's Robotics Institute have built modular robots specially designed to crawl inside natural gas pipelines, map their layout, detect damage, and apply protective resin coatings. The projects aims to address aging and deteriorating gas distribution systems critical for powering over 75 million American homes.

"We're going to see pipes bursting more often...because we're not going to be proactive in addressing our pipe infrastructure," says CMU professor Howie Choset, who leads the effort sponsored by the Department of Energy's Advanced Research Projects Agency-Energy (ARPA-E).

The robots sit on wheels straddling the pipe and inch along while gathering high-resolution imagery. Custom optical sensors map pipe geometry and textures since conventional solutions like lidar struggle in the cramped diameters.

So far, the system has charted over 9 miles of piping in 8 hours. It can also apply reinforcing resin to breach sites at around 1.8 miles per 8 hours. "What makes ours different is they're very strong," Choset explains. "They have potential to do some kind of repair."

But a recent shift in ARPA-E program management means further repair development funding remains uncertain. "This is an important problem," Choset urges. "In 50 years, we're not going to have pipes." He's cobbling together resources to continue the work.

The team also collects images to train AI for autonomous damage detection. "We have a pipe network, we have a time-varying pipe network, and an artificial intelligence can help inspectors look for problems while they're still small," Choset notes.

Deliberate modularity lets technicians customize robots by swapping components like wheels, payloads, and batteries based on needs. Choset believes no single configuration is optimal.

The modular flexibility also speeds development. "Instead of a new system every time, you just develop a module," he says.

Choset hopes to commercialize the technology this year through his company JP Robotics. He sees potential across applications involving navigation and manipulation in confined spaces like nuclear plants or search and rescue.

"I'm interested in robots in general," Choset concludes. "I just find maneuvering, sensing, and predicting in tight spaces to be a very interesting problem."