hide announcement WIN your very own meteorite! Subscribe, gift or renew a subscription to Cosmos and automatically go into the draw – Shop now!

The robots that could colonise Mars

Swarming machines with the termite's instincts may act as pathfinders and builders of a Martian colony. Elizabeth Finkel reports. 

240214 marsrobots h.jpg?ixlib=rails 2.1
Robots inspired by termites may build our first colonies on the red planet.
UIG via Getty Images

Robots were the first explorers on Mars. They will probably build our first dwellings there, too. In chilly Chicago at the annual meeting of the American Association for the Advancement of Science (AAAS) last week, an audience of eager journalists was treated to a demonstration of just how the robots might do it. Like termites, it turns out.

As the eager scrum looked on amidst flashing cameras, robot designer and builder Kirstin Hagelskjaer Petersen patiently cajoled her error-prone bot, about the size and shape of a Tonka truck, through its demo. “The flashes are highly disruptive,” the gentle, intense PhD student from the School of Engineering at Harvard University remarked. Nevertheless, by the briefing’s end the bot had successfully erected a three-step staircase, one foam brick at a time.

The climbing robots at work. As with all swarm intelligence, a small repertoire of programmed behaviour allows the bots to work together to build complex structures. – Self-Organizing Systems Research Lab, Harvard University

This simple feat is just the beginning of what these bots can do. As a paper in Science reported on 13 February, three of these bots have worked together to build a trident-shaped single storey structure in 30 minutes. The structure itself isn’t too remarkable; remote-controlled robots have built complex structures before. What impresses is that the bots are completely self-directed. Each acts autonomously, yet through “swarm intelligence” they collectively construct something complex.

In the human world, building is a top-down affair. An architect designs it down to the last detail, then instructs the builder at each step who in turn instructs specialist tradespeople. The process can achieve the magnificence of the Sydney Opera house or Singapore’s “flying boat” Marina Bay Sands hotel, but requires an elaborate management hierarchy. Take away the architect or foreman and construction stops.

Termites, bees and other creatures with swarm intelligences do it differently. Each individual reads local cues, such as a scent trail, and responds by following a simple set of genetically “programmed” instructions. Collectively their feats are impressive. Colonies of millions of Namibian termites, for instance, build mounds more than five metres high, replete with many complex, air-conditioned chambers.

Autonomous, swarm-intelligent robots might one day build in sites hostile to humans, such as disaster zones or colonies on Mars, says Justin Werfel from Harvard’s Wyss Institute for Biologically Inspired Engineering, who, along with Petersen, designed the bot’s coding algorithms. The more bots, the faster the structures rise but should a meteorite strike or other catastrophe wipe some of them out, just like termites, the rest will carry on.

Trying to mimic termite code in robotic systems had proved elusive. So Werfel set himself a mission: to work backwards from complex structures to find a set of simple rules that would produce it.

Across the Atlantic in Denmark, electro-technical engineer Petersen, who’d grown up building with Lego, had also become fascinated by termites. She had started building prototype termite-inspired robots, but their algorithms weren’t up to the task. When the two engineers’ paths finally crossed, it was clear they were a perfect match. Their efforts were supervised by Radhika Nagpal, a professor at Harvard’s School of Engineering and Applied Sciences and also the Wyss Institute.

Four years later and their clumsy-looking insect-like critters were ready to build a foam brick staircase at AAAS. Each white 20 cm by 20 cm by 4 cm brick was marked with a black cross, the better for the bot’s infra-red eyes to see them. The bots also have ultra-sound detecting “ears” to register each other’s presence; pressure detectors in the form of squishable buttons; and angle detectors so they know where they are during the climb. Each bot is also equipped with a front loader to grab and place the brick. And what allows them to climb is their star-like “whegs”, a fusion of the words wheels and legs, which is just what they are. Legs are good at climbing but pretty average at running; wheels the opposite. Designing just the right whegs for the bots was a painstaking effort by Petersen, says Werfel.

Yet for all their achievements, the bots are very simple, simpler than termites. All they can do is sense bricks and the robots next to them, and modify their behaviour accordingly – a repertoire limited to moving backwards or forwards, turning in place, climbing one step up or down and picking up, carrying and depositing bricks.

“This system is extremely elegant,” noted Judith Korb from the University of Freiburg in Science. “It is robust to failure because of its decentralisation, and is very flexible in its adjustments.”

Simplicity and reliability are the words Petersen and Werfel repeat over and over, but in practice they are hard to achieve. As Peterson performed her demonstration, she gently nudged the ungainly bot when it went awry. It was not cheating, she explained. “They are designed to correct their errors.” Her actions were not unlike those of a mother picking up her clumsy toddler. “It’s true. I can’t help transferring emotions to these guys,” Petersen admitted. She recalled her days at Pasadena’s Jet Propulsion Lab when she worked on the prototype for the Mars Rover. “The Mars Rover was definitely a ‘she’.”

Though the endearing bots were modelled on termites, Werfel and Petersen say they still don’t fully understand how termites build their impressive castles from simple rules. But that doesn’t mean human engineers might not yet better them. As Korb suggests, “The algorithms of Werfel and colleagues might also show us whether termites could still 'learn' from humans.”

Ella finkel twic.jpg?ixlib=rails 2.1
Elizabeth Finkel is editor-in-chief of Cosmos.
Latest Stories
MoreMore Articles