Imagine a team of workers that can tirelessly build and rebuild complicated structures even under daunting and dangerous conditions. They exist — they’re called termites. Now, inspired by these mound-building insects, Harvard University scientists have created a mini-swarm of surprisingly simpleminded robots that can work together to construct buildings much larger than themselves.
The findings, described in the journal Science, present an important step toward designing robots that may one day be able to build research facilities in the deep ocean, buildings on Mars or even levies at a flood zone during an emergency — jobs that are far too hazardous or expensive for human workers to do.
Termites can build high, skyscraper-like mounds up to about two stories high. They achieve this architectural feat without a leader or a blueprint to tell them how and where and what to build. In fact, the insects don’t seem to take any orders — they just operate on a simple set of rules that somehow result in these complex structures. It’s the same idea behind “swarm intelligence” — how hives of bees or colonies of ants can act as a complex organism together even though they’re each functioning with a very basic set of rules.
“How can such tiny insects, each less than 1 centimeter in size and equipped only with a simple brain, construct air-conditioned buildings up to 500 times their size?” Judith Korb of the University of Freiburg in Germany wrote in a commentary on the study.
Termites have benefited from eons of evolution to tweak and refine the simple rules they use to get the job done, Korb said.
“A termite mound’s architecture can determine the success of a colony,” Korb wrote. “Mounds that are better adapted to local environments will, as a rule, have more offspring; thus, improved building rules that are genetically encoded will spread over time through a population.”
To see if they could harness this ability, scientists at Harvard’s Wyss Institute for Biologically Inspired Engineering created robots that are remarkably simple — they have three actuators to move, sonar to sense and a few other capabilities. They can move forward, backward, turn in a circle, climb up a single step at a time, pick up an object and put it down, and use sonar to sense its peers and the structure it’s building.
“The inspiration from termites was really high level — general principles about the way they work,” said lead author Justin Werfel, a Harvard computer scientist. “The fact that they’re decentralized and independent, the fact that they have limited and local sensing, the fact that they’re reactive rather than planning ahead, and the fact that they coordinate using the environment rather than direct communication.”
The scientists programmed the robots with two simple sets of rules. One is a more general set of rules — for example, what safety checks to make before setting down a brick in a particular spot — and the second involves a set of traffic rules specifically designed for the individual blueprint of whatever building they were charged with setting up.
This was a clever trick: A lot of swarm robot research first builds the robots, equips them with a set of rules and tests whether those rules work. But it’s hard to get a complex design by building from the bottom up — the results can be unpredictable.
Instead, the computer scientists created blueprints for the buildings they designed and then worked backward to translate it into simple, specialized traffic rules that the robots could use.
These robots are nowhere near ready for prime time yet; their pinwheel-shaped wheels are good for climbing but not so good for turning. And they can build only certain types of structures.
Right now, they’re working with specially shaped lightweight bricks designed for this very simple machine.
“If you make the robots more complicated, then you can use more simple building materials,” Werfel said.