A shoe-sized robot, shaped like a VW Beetle and built by a 3D printer, scuttles in circles on a Harvard lab bench. Its hooked wheels, good for climbing and grasping, also let it trundle on the flat. As I watch, it scoops a styrofoam block on to its back and then scrabbles across a layer of already deposited blocks to flip the new one into place. An impressive feat – especially given that it does this without human control, using simple rules about its environment to build a whole structure.
The robot is making a tower – like a termite might.
“If you want to build underwater, if you want to build a Mars base, it’s going to be very difficult, dangerous and expensive to send people,” says Justin Werfel of the Wyss Institute for Biologically Inspired Engineering at Harvard University. “But if you could send a team of robots to go build the habitat as the first step – that’s the really long-term vision.”
To spur his swarm into action, Werfel gives the robots a mathematical model of the structure to be built, say, a pyramid. Each robot uses that model to calculate where it will place the next block it picks up, moving on to another spot if its planned drop-off has already been completed by another bot. It uses nothing but basic ultrasound and infrared sensors, as well as an internal accelerometer, to figure out how many blocks it has climbed, and where it is in relation to the structure it is building.
As well as pyramids, the bots can build castles and towers.
Build by numbers
Neri Oxman of the Media Lab at the Massachusetts Institute of Technology, who works on robotic architecture, says the concept is very promising. “This work promotes a truly decentralised construction system offering robust and customised designs,” she says. “It paves the way for emergent design based on environmental sensing and represents an important step towards enabling the shift from swarm-based construction to swarm-based design.”
The robots – part of Harvard’s termite-inspired TERMES project – are never going to build anything outside an academic setting, but they show that controlling robots using simple, distributed rules does work. “You have a simple, cheap, expendable robot, and then you throw a bunch of them at a system and you don’t care if they break,” says Kirstin Petersen, who designed and built the robots. “You can get really far with simple robots.”
Petersen and Werfel are not the only roboticists who see value in the swarms. The Laboratory of Intelligent systems at the Swiss Federal Institute of Technology in Lausanne has developed a swarm of small aircraft that communicate, position themselves and find targets using nothing but sound, without any central control system. Instead, the Swiss researchers give each flying robot the ability to make decisions based on its local conditions. At Heriot-Watt University in Edinburgh, UK, roboticists are working on a swarm which can repair damaged coral reefs.
In the short term, Werfel says TERMES-like robots could be used to build levees out of sandbags, working through night and day without human help to build flood protections. “The scenarios where you want to use robots rather than humans are described by the three Ds – dirty, dangerous and dull,” he says.
The group has already used a modified version of the TERMES robots to drag and stack bags of rice to make a wall.
Journal Reference: Science, DOI: 10.1126/science.1245842
19:00 13 February 2014 by Hal Hodson
For similar stories, visit the Robots Topic Guide