Inspired by Termites, Bots Build Complex Structures With Simple Rules

No foreman necessary: These robots sense changes in their local environment and use those cues to build without detailed supervision.
 
Guided by simple rules, bio-inspired robots build a structure without complicated communication or coordination. | Courtesy of Science/AAAS

CHICAGO Robots inspired by the collective construction sites of termites can build complex block structures using just a few simple rules, researchers reported at the AAAS Annual Meeting.

Unlike humans, who require a high-level blueprint to build something complicated, termites can build complex mounds hundreds of times their size without a detailed plan. They take simple cues from each other and their environment to know where to lay the next clump of dirt, and ultimately, to build a structure that suits their surroundings.

Justin Werfel of the Wyss Institute for Biologically Inspired Engineering and colleagues demonstrated at a press briefing how the robots can be programmed to follow the same kinds of cues to build a structure requested by human. A paper published in the 14 February issue of Science also described this study.

"Termites are what inspired this whole research topic for us," said Werfel. "We learned the incredible things these tiny insects can build and said: 'Fantastic. Now how do we create and program robots that work in similar ways but build what humans want?'"

Over the long term, at least a few decades, Werfel envisions using teams of these robots to build full-scale structures in difficult or dangerous settings such as underwater habitats or Martian plains. "While that's likely a long way out," he said, "a shorter-term application [on the scale of years and not decades] could be something like building levees out of sandbags for flood protection."

The use of local information to coordinate behavior is called stigmergy. Werfel and colleagues leveraged stigmergy to design algorithms that reflect termite behavior and then implemented these algorithms in their robots. 

Their bots need only the ability to sense a brick or another bot nearby to make their next move. Equipped with sensors, they move along a grid, lifting and depositing bricks. If they sense a brick in their path, they carry their cargo to the next open spot.

And they do all this without a detailed plan or centralized communication; instead, the bots are programmed with just a few simple rules.

"There are two kinds of rules," Werfel explained. "The rules that are the same for any structure the robots build, and the 'traffic laws' that correspond to the specific structure. The [traffic laws] tell robots at any site which sites they're allowed to go to next. Traffic can only flow in one direction between any two adjacent sites, which keeps a flow of robots and material moving through the structure."

Werfel further explained why the robots won't place bricks just anywhere. "If they built carelessly, it would be easy for them to build in a way where they got stuck," he said. "The safety checks involve a robot looking at the sites immediately around itself, paying attention to where the bricks already are and where others are supposed to be, and making sure certain conditions in that local area are satisfied."

Science -co-author Kirstin Petersen, an academic fellow at the Self-Organizing Systems Research Lab at Harvard University, showed off one of the robots as it built a small block staircase at the news briefing. Spinning its claw-like wheels, the roughly 4 x 6-inch robot searched for foam blocks, hoisted them onto its back and climbed the growing staircase to deposit its load in the correct place. The only hitch came when a photographer's flash confused one of the robot's sensors.

Kirstin Petersen demonstrates termite-inspired robots at an Annual Meeting news briefing | AAAS/Janel Kiley

The robots make little errors all the time as they climb and build, but they are programmed to use this feedback, Petersen said. "We're not trying to make it perfect; we're trying to make it aware of its errors and try to correct them."

Though each robot "knows" only simple rules, like when to put a brick down, turn around, or climb one step higher, the robots together exhibit intelligent behavior in completing structures.

Independent robots with decentralized control have numerous advantages, Werfel explained. "Individual robots can break down but the rest can carry on. There's no one critical element that brings everything down if one fails."

Such systems are also scalable, he said. "For a bigger job, you can just add more robots (even mid-job) without needing to change how they're programmed." By contrast, a robotic system with a centralized controller could create a bottleneck, with a limit of how much it could coordinate as new robots came onto the scene.

Although the researchers drew their original builder rules from the literature on termite building, they recently contacted J. Scott Turner, a physiologist at SUNY College of Environmental Science & Forestry and an expert in termite mounds, to discuss their next rounds of robot refinement.

"We really want to go back and look at the termites, and study the details of what the termites are doing, to figure out what program the termites are running and figure out how to connect that to the structures that they build," Werfel explained.