Jun 14, 2021 7:00 AM

How Do You Make a Robot Walk on Mars? It’s a Steep Challenge

Meet SpaceBok, a little, four-legged machine that’s taking the first steps toward walking on the Red Planet’s brutal terrain.

From the Sojourner rover, which landed on Mars in 1997, to Perseverance, which touched down in February, the robots of the Red Planet share a defining feature: wheels. Rolling is far more stable and energy efficient than walking, which even robots on Earth still struggle to master. After all, NASA would hate for its very expensive Martian explorer to topple over and flail around like a turtle on its back.

The problem with wheels, though, is that they limit where rovers can go: To explore complicated Martian terra like steep hills, you need the kinds of legs that evolution gave animals on Earth. So a team of scientists from ETH Zurich in Switzerland and the Max Planck Institute for Solar System Research in Germany have been playing around with a small quadrupedal robot called SpaceBok, designed to mimic an antelope known as a springbok.

True to its name, a real-life springbok bounces around the deserts of Africa, perhaps to confuse predators. The original concept for the robot, which was introduced in 2018, was actually for it to jump on the surface of the moon, as astronauts have done to locomote in the weak lunar gravity. That may work on our satellite, where the landscape is relatively flat, but on Mars it’s probably too risky given the complex terrain—which is full of sand, rocks, and steep slopes. So now the researchers are modifying its limbs and gaits to see if it might be able to handle more brutal landscapes.

In these new experiments, the team programmed SpaceBok with more traditional, less springy gaits. Specifically, the researchers wanted to compare two kinds: a “static” gait, in which at least three limbs are making contact with the ground at any given time, and a “dynamic” one, in which more than one limb can leave the ground at once. The former is more methodical, but the latter is more efficient because it allows the robot to move faster.

The researchers also outfitted versions of SpaceBok with two kinds of feet: point and planar. The point feet have a small surface area, kind of like the hoof of an actual springbok. The planar feet, by contrast, are actually flat swiveling circles, which bend at an angle when the foot makes contact with the ground. Think of these more like snowshoes than hooves. Or really, they’re like snowshoes with cleats, since they’re studded with projections that help the foot grip the ground.

Notice the large surface area of the planar feet

Courtesy of Hendrik Kolvenbach/ETH Zurich

Once the researchers had different configurations of gaits and feet they could use to customize the robot, they set it loose in a giant tilted sandbox loaded with material that approximates the soil found on Mars. That way, they could test whether any of those configurations allowed the robot to get up a 25-degree plane. By monitoring the robot’s energy usage, they could quantify how efficient each of the configurations of gaits and feet were.

In a new preprint describing the work, which has been accepted for publication in the journal Field Robotics, they showed that the machine can deftly and efficiently climb a simulated Martian hill without tumbling down it. “We wanted to show that these dynamically working systems nowadays, they can actually walk on the Martian sand,” says ETH Zurich roboticist Hendrik Kolvenbach, the study’s lead author. “This is a technology that has a lot of potential now for the future.”

Interestingly, the robot got up the hill just fine using both the flat feet and the pointed ones. The flat version allowed the robot to rest on top of the sand. The pointy version would instead sink, providing a sort of anchor. “One of the surprise findings was that the point feet were not performing so bad on this particular slope, because of that high sinkage,” says Kolvenbach. “Basically, they provide quite a stable stance.”

Well, at least that was true of this simulated Mars soil. On the actual Red Planet, there may be rocks hidden in the sand—the robot could take a tumble if it caught one of those. Buried rocks are particularly challenging obstacles because the robot wouldn’t be able to detect them with its camera. It wouldn’t know it had a problem until after it fell over. (Researchers can equip SpaceBok with a camera for autonomous navigation, but for these experiments it was walking blind.) In the case of rocky terrain covered with sand, a robot with point feet would be more likely to strike hidden stones. The flat foot, the team found, made the robot slower, but they think its shape makes it more likely to safely pass over buried impediments.

The robot attempts a static gait with planar feet

Courtesy of RUAG Space/ETH Zurich

But the flat feet also had some drawbacks. Because the sandbox was angled, material slippage was another big challenge. Think about what happens when you clamber up a dune and you get those little sand avalanches around your feet. It takes more energy to get up that slope if the sand is constantly moving underneath you—you’re fighting both the incline and the debris. And for SpaceBok, since the flat foot caused more of a surface disturbance, it increased slippage, while the point feet, which sank into the ground like stakes, minimized it. “The flat foot was actually performing worse, energetically, because we had more slippage,” says Kolvenbach.

The ideal design is probably somewhere in between the two, more like the foot of a camel—not preposterously wide like a snowshoe, but not too skinny, like an antelope’s hoof. “There's a sweet spot,” Kolvenbach says. “I think you definitely need this increased surface area compared to the point foot, because you really want to avoid these high sinkage events, where you might not be able to get out again anymore. On the other hand, you don't necessarily want to have these huge flat feet.” In the future, Kolvenbach adds, they might even be able to design a foot for SpaceBok that modifies its surface area in real time to adapt to different kinds of soils.

A four-legged robot would need a similar flexibility in its gait if it were to walk the real Red Planet. The robot is safer while using static locomotion, in which it always keeps at least three legs on the ground, than when using dynamic locomotion, which more closely approximates how four-legged animals move. But it turned out that the static locomotion was actually less efficient for SpaceBok as it tried to summit the slope. “You are limited by the speed of one leg that pushes you forward,” says Kolvenbach. “Whereas in a dynamic motion, you have at least two feet that push you forward. So you're just becoming much faster. And overall, because you also need some energy to suspend the weight of the robot, it will allow you to save some energy.”

And here's a dynamic gait with point feet

Courtesy of RUAG Space/ETH Zurich

So a future SpaceBok would need to be able to switch its gait, along with the shape of its feet. On the plains, it can use a dynamic gait to move faster and save energy while getting from point A to point B. When it’s trying to get up a particularly gnarly hill, it might switch to a static gait to walk more safely, sacrificing energy for not plummeting down the slope.

A pathfinding strategy is critical, too. In these experiments, SpaceBok was equipped with an algorithm that monitored its energy usage to automatically determine the most efficient path to take. This produced an “emergent” behavior, in which the robot opted for zig-zag switchbacks as it climbed, instead of scurrying up the hill head-on, which would have been more of a struggle and therefore more of a power suck.

This intimate interplay between a robot’s hardware, software, and surrounding environment is part of a larger trend in “embodied” robotics, says roboticist Tønnes Nygaard of the Norwegian Defence Research Establishment, who studies quadrupedal locomotion. With embodied robotics, engineers are training machines to adapt to difficult terrain, which human bodies do so easily. We don’t think twice about how we should coordinate that dance of muscles. Ideally, a robot walking on Mars would be similarly adaptable, especially since it would require a high degree of autonomy, thanks to the communications delay from Earth.

The prospect of a robot that’s not limited by its wheels is exciting to researchers, who have a keen interest in investigating sandy or steep terrain. “We are often interested in these areas, particularly craters, where we know there were once ancient lakes,” says planetary scientist Mariah Baker of the National Air and Space Museum, who has worked on the Insight lander, as well as the Curiosity and Perseverance missions. That’s because where there was once flowing water, there may well have been life. “As we kind of establish new ways of traversing and exploring, possibly with these new kinds of robots, it might open up parts of the planet that we haven't been able to explore before,” she says.

A descendant of SpaceBok, then, may one day go where no rover has gone before to search for Martian life, joining the new Mars helicopter in a diversifying army of science machines. “Legged robots might not replace wheeled robots in space,” says Nygard, “but they could definitely bring a valuable contribution and take an important role in the team.”