Two years ago, Alice Agogino, a UC-Berkeley mechanical engineering professor, was working on a contract to build exploratory robots for NASA Ames. She had been recruited to help design what would eventually become a fleet of mobile, ultra-impact-resistant, remote-sensing robots that could protect sensitive scientific equipment during a drop from orbit onto the surface of a moon—specifically Titan, an ice-covered moon of Saturn.

But then she read a report that brought her research back down to Earth.

The report, from the International Red Cross and Crescent, suggested that a generous portion of casualties among first responders—emergency workers tasked with initial disaster management—could be linked to poor situational awareness on the ground.

Suddenly, Agogino recognized the potential for a brand-new use for her robots.

They Came from Outer Space

As Agogino read the report, she knew what she was creating could be the wave of the future on our own planet. Situational awareness was precisely the goal of the NASA bots, whose every feature was designed to protect and operate the most advanced sensory equipment available.

Everything that made the squishy bots the perfect space recon agents—their autonomous sensing power, remote control capabilities, and unprecedented impact durability—would serve them equally well as members of the advanced guard of Earth-bound disaster responders.

Yellow, Rock, Metal,
Lee-Huang Chen
The robot.

To realize her vision, Agogino and two members from the NASA team incorporated Squishy Robotics, a startup out of UC Berkeley’s SkyDeck accelerator. The starting model was a static version of the mobile robot Agogino had been working on for NASA, which was about the size and shape of a geodesic soccer ball.

Once the lab was staffed and funded—in part by the same NASA Ames grant that had started it all—there was just one thing Agogino had to do: learn everything there is to know about disaster response.

The Ecology of Disasters

Over the course of 18 months of research, the Squishy Robotics team conducted over 200 interviews with first responders across the United States.

“Everything was surprising,” Agogino says of the experience. “There’s a whole ecology around first responders. It’s not only the fire department—it turns out it’s Homeland Security, the military, and even utilities employees. A lot of fires are being started by faulty industrial electronic equipment.”

The firefighters, federal officers, and utilities workers invited Agogino and her research team to explore the complexity of their work by going over work plans, demonstrating the use of their equipment, and pointing out the problems they face daily as they try to minimize the damage of a huge range of possible disasters.

The common refrain, however, was the challenge of situational awareness: How could they get vital information about safety hazards at the disaster site before they could pose a threat to rescuers or bystanders?

The lack of foresight about conditions on the ground is much worse than an inconvenience. In 2005, the worst chemical hazard disaster in U.S. history took place following a train collision in Graniteville, South Carolina. While rescue workers were suiting up nearby to intervene, they had no way to know that poisonous chlorine gas was already spreading from a damaged tanker on one of the trains into the surrounding residential area.

With no early warning signs to begin evacuation procedures, the industrial disaster management team couldn’t issue the order to evacuate until the chlorine had formed an enormous cloud of airborne poison in the low-lying valley of the surrounding area, when some of the nine fatalities and 631 chlorine gas injuries had already occurred. The event is still used as a training scenario for first responders.

More recent disasters have reiterated the deadly cost of a lack of intel on the ground.

One of the three fire departments now partnering with Squishy Robotics to flight-test the bots is the Houston Fire Department, which was responsible for addressing the unexpected industrial side effects of Hurricane Harvey’s massive flooding in 2017.

“You wouldn’t think a hurricane would cause electrical issues,” Agogino says, “but those power outages shut down a chemical plant near Houston, causing materials like peroxides to heat up and explode.”

The plant, multinational chemical manufacturer Arkema Inc., was later sued by first responders who sought medical treatment after exposure to the toxins, Mother Jones reported.

Robots at the Scene of Catastrophe

Few in the engineering world understand this challenge better than Robin Murphy, a Texas A&M professor of computer science and engineering and co-founder of the field of disaster robotics.

As vice president of the Center for Robot-Assisted Search and Rescue (CRASAR), Murphy has had a hand in robotic search-and-rescue operations for 28 disasters, ranging from the World Trade Center attacks to mudslides, hurricanes, mine collapses, floods, volcanoes, and nuclear incidents—all of which qualify as CBRNE (chemical, biological, radiological, nuclear, and explosive threat) events, according to Homeland Security.

Murphy’s work covers disaster robotics theory, CBRNE hazards, and rescue operations, which could all eventually benefit from situational awareness technology like squishy robots, she says.

“I think it’s exciting. There’s lots of possibilities [for this approach],” says Murphy. “Ground sensors are particularly useful for CBRNE environments. It’s something we wish we had had for the La Conchita mudslides.”

Cartoon, Illustration, Fiction, Fictional character, Art, Animation, Games, Graphic design, Drawing,
Alice Agogino/Squishy Robotics
A rendering of how the robots could be deployed in search-and-rescue operations.

The deployment of CRASAR robots at those mudslides in La Conchita, California, which occurred in the same month as the Graniteville train disaster, was largely considered a failure, in part because of a lack of training data with which to prepare the unmanned mobile CRASAR robots sent into demolished homes to search for survivors. Because the robots could not sense the seriousness of a threat of further slides, they had to be removed from the site within minutes of deployment.

Throughout her career in disaster robotics, Murphy has observed a recurring theme of complex and unpredictable terrain defying even the most ambitious technological interventions.

“The big problem is getting enough power to get a signal back, to return that data [from the ground sensors],” says Murphy. “The wireless isn’t very good. When you start dropping these things into rubble, or even a chemical train derailment, just one piece of rebar metal can block the wireless signal.”

Other challenges include mobility on shifting terrain for the next generation of squishy robots, as well as the severe limitations of building something small and light enough to fly without losing valuable equipment capacity.

“Real situational awareness requires advanced cameras, GPS, chemical and radiological sensors, proprioceptive sensors—not to mention processors to turn all that data into a usable signal,” Murphy says. “These things start adding weight, cost, and energy consumption, all of which make the units harder to build and operate.”

Even more frustrating? The surprising vulnerability of the typical robotics materials to the harsh conditions of disasters.

“Biological systems like arms and legs are cheap,” Murphy dryly says, noting the relative hardiness of human mobility systems when compared to their mechanical counterparts. Robots, on the other hand, are “susceptible to dust, corrosion, and water damage”—to say nothing of the impact of a 400-foot drop to the Earth’s surface.

Bringing the Bots Back to Earth

That drop was the one engineering challenge that the Squishy team had already mastered when Agogino began to interview first responders. For the NASA contract, Agogino and other engineers had turned to the wisdom of 20th-century inventor and architect R. Buckminster Fuller. One of his most lasting contributions to design was the concept of tensegrity, a portmanteau of tensional integrity.

Vehicle,
Alice Agogino/Squishy Robotics
The squishy robot falling to Earth.

Tensegrity structures are marked by their high strength-to-weight ratios and their ability to distribute forces delivered to one section of the object across the entire structure.

“Technically, it means all forces are axial,” says Agogino, adding that impacts would be translated to structural compression rather than causing twisting or breakage: “That’s what makes the robots squishy.”

Agogino says that tensegrity structures have appeared in sculptures, artwork, and even buildings, like the famous geodesic dome popularized in the U.S. by Fuller, but Squishy Robotics represents the first formal effort to incorporate the tensegrity principle into a robotics project.

With the help of the first responders who shared their knowledge and experience with the research team, Squishy Robotics developed a model that expert rescuers would be enthusiastic about using in their work—even when that meant deviating from the engineers’ contrary instincts.

Glasses, Eyewear, Electric blue, Vision care, Portrait, Official, Portrait photography, Smile,
Alice Agino/Squishy Robotics
UC-Berkeley mechanical engineering professor Alice Agogino.

“They’re not autonomous robots, at the request of the first responders, who don’t trust autonomous systems in these dangerous environments,” Agogino says. The first responders also requested live camera feeds so they can see what the robot sees in real time, a feature now incorporated into the standard squishy robot.

A Squishier Future

Field testing for these stationary sensor robots has already begun, thanks to partnerships with the Houston, Alameda, and LA County fire departments, but that doesn’t mean the Squishy team are getting complacent. The first-responder interviews yielded a slew of new avenues for future iterations of the robots, several of which are already being tested in the lab.

“We’re working on putting rotors on the robot to make them low-to-the-ground drones, which can maneuver throughout buildings,” Agogino says. “And we’ve had a request to make them float [for use in water rescues and flood incidents].”

Translating the accomplishments of the stationary squishy bots to mobile versions will be its own challenge. “The first mobile robot we built for NASA could only withstand a five-foot drop,” Agogino recalls.

But part of the goal of the fire department partnerships is to find out how to apply successful practices to the complex machinery of a mobile squishy bot, which would have to include both sensor equipment and space-ready mobility systems like cold-gas thrusters. “We are learning from the stationary robot how to design and build the mobile version so it can be dropped from higher,” says Agogino.

A moon landing may be as distant as two decades away, Dr. Agogino says, but in the meantime, the genius of the robots’ design could potentially save hundreds of lives. And there’s no denying that the bots themselves are compelling creations with the ability to increase people’s comfort with working alongside robots in the field.

“You can’t help but feel affection [for the robots]—they appeal to a lot of people,” Agogino says. “They’re not scary like some security robots or defense machines. We embrace the fact that we are lightweight and can work around humans without hurting them.

Headshot of Lynne Peskoe-Yang
Lynne Peskoe-Yang

Lynne Peskoe-Yang is a science writer and researcher in New York. Her reporting on civil engineering, machine learning, and artificial intelligence has appeared in MarketplaceIEEE SpectrumRewire.org, and Sludge; her essays on science and language live over at Popula