We think of termites as agents of destruction. Here in North America, the little buggers chew through untreated wood and give homeowners headaches. But in the Southern Hemisphere, colonies of termites—each less than a centimeter long—collaborate to build complex mud mounds rising up to 25 feet high.
The bugs do this without a blueprint or any supervision. There’s no tablet-toting termite foreman directing the workers. Nor is there specialization among workers. Each termite has the same, limited set of skills. As individuals, they’re expendable: If a human steps on some or an aardvark slurps up a bunch mid-task, the others take up the slack. But because there are a million of them to a colony, the insects—sharing a collective goal and a few instinctive rules—eventually get the job done. (It can take a year, or years, just like our own skyscrapers.)
Harvard’s Kirstin Petersen and Justin Werfel flank a termite-built mud mound in Namibia. (Photo by Radhika Nagpal)
To a few Harvard scientists, these massive mound projects begged the question: Could the same principles be applied to robotic builders?
Justin Werfel, Radhika Nagpal, and Kirstin Petersen form a joint team from Harvard University’s Wyss Institute for Biologically Inspired Engineering and School of Engineering and Applied Sciences. Their ongoing research is so far most notable for the Termes experiment, in which small, simple robots worked independently to build proportionally large structures, without a human or even a smart computer coordinating their efforts—at least, not in the way we might think of it.
The experiment made waves after the journal Science published the results in February 2014. Since then, the Harvard researchers have been traveling to Namibia to learn more about the termites that inspired the project. Werfel, who has also authored a book chapter on the ecology of Fraggle Rock (remember the Doozers?), walked us through the Termes project.
Of whegs and stigmergy
The Termes system was realized both in computer simulations and in a physical lab setting. “In the simulations, you’ve got dozens of robots building enormous skyscrapers flawlessly,” said Werfel. In the latter—i.e., the hardware—came the true test. Built by Petersen, the three Termes robots measure about seven inches long and 4.3 inches wide. Each is equipped with infrared and ultrasound sensors, a clawed arm capable of lifting in the manner of a front-loader, and a kind of tail to aid in carrying cargo.
Meet the Termes robots. (Photo by Eliza Grinnell)
The robots move about on “whegs”—combination wheel-legs, in a triskelion shape (like on the Manx flag). This is to help the robots climb up and over the blocks they’ve already laid. Though the Harvard team didn’t invent whegs, Werfel recalls learning about wheels with protruding sticks for enhanced mobility when, as a child, he saw the following TV commercial:
As with termite workers, no one Termes robot has a special skill or role. Each is capable of the same few simple tasks—walking and climbing, picking things up and putting them down.
The bots’ building materials are interlocking foam blocks, shaped somewhat like electric beer coasters—the kind that flash and hum when your table is ready at the steakhouse. The blocks fasten together magnetically, so the robots don’t have to be precise when stacking them. They’re also notched at the edges, aiding the robots in gaining a foothold as they climb up or walk across them.
The Termes robots stack and climb up specially designed blocks. (Courtesy of the Harvard University Self-organizing Systems Research Group)
Werfel’s team fed the robots the design plan for a structure, such as a staircase or a pyramid—which are, not coincidentally, well suited to the robots’ climbing ability. The bots were also coded with a set of “traffic laws” to avoid crashes.
From there, the robots were on their own. To make the system work the way termite colonies do, Werfel said, “the robots need to be independent; they need to have knowledge that only they can sense themselves; they’re going to be building large-scale [structures]; they’re going to be climbing over things they build in order to get to higher places they couldn’t otherwise reach . . . and they’ll be coordinating through indirect communication.”
That indirect communication occurs “via the joint manipulation of a shared environment,” as Werfel wrote in the Science article. In other words, as the robots work in parallel, moving blocks around, they leave one another cues (and in turn pick up on cues) as to what should be done next. This form of communication by implicit coordination is called “stigmergy.”
In this decentralized approach, there’s no prescribed order in which the robots must stack the blocks, just a prescribed outcome. It’s up to the robots how they get the job done. (Not that they plan this in advance, since they’re simply behaving reactively—remember: stigmergy.)
Images from the simulation side of the Termes project. (Courtesy of the Harvard University Self-organizing Systems Research Group.)
The advantage with such a system of several unsophisticated robots is that there’s no single failure point. If one of the robots falls on its back and can’t get up, the other two keep working and finish the job. That said, little failures in the hardware sometimes added up to an overall failure, as when a robot fell in a spot where the others couldn’t pass it. “And we didn’t have a tow truck,” said Werfel. “Well, we did; it was Kirstin. In a real system, you’d want a tow truck.”
Nevertheless, the robots ultimately succeeded in building a structure ten blocks tall. In Werfel’s terms, they proved that an emergent outcome can be engineered into a system of low-level independent agents. Check out the time-lapsed video, below, showing the Termes robots at work:
Out of the lab
So will giant Termes-style autonomous robots build high-rise hospitals or luxury condo towers in the middle of dense cities with no supervision? Not likely in our lifetimes. However, we might well see such robots in action during a flood, piling sandbags to build emergency levees. Or in another disaster scenario, Werfel suggested, the robots could enter an area struck by an earthquake to shore up a shaky building. “You don’t want to send in people because the building could actually collapse while you’re trying to reinforce it,” he said. After all, the robots are expendable, while people aren’t.
In the more distant future, Werfel envisions similar robots being used to build Martian homes or deep-sea research stations in advance of human explorers’ arrival, or in other situations when it could be considered highly impractical or prohibitively dangerous to employ us flesh-and-blood types.
Paradoxically, if such high-tech facilities can be erected without the loss of a single human life or limb, we’ll have the lowly termite to thank for the inspiration.
This post was a collaboration between Suffolk Construction’s Content Writer Patrick L. Kennedy and former Suffolk Construction Marketing Intern Jen Howard. If you have questions, Patrick can be reached at PKennedy@suffolk.com. You can connect with him on LinkedIn here or follow him on Twitter at @PK_Build_Smart. You can also connect with Jen on LinkedIn here.