AS CARS HERE on Earth begin to drive themselves and robots autonomously roam sidewalks delivering food and nearly running over dogs, over on Mars, the Curiosity rover very much remains a remotely piloted vehicle. That’s part technical restraints, part design: Curiosity is up there to do science, so it must follow the commands of scientists here on Earth.
Still, that doesn’t mean automation can’t lend a hand. Back in May of 2016, NASA began using an autonomous targeting system, called AEGIS, that allows the Curiosity rover’s cameras to automatically detect preferable terrain to sample. Over the next year, the system was able to eyeball and automatically identify suitable rocks with very high accuracy, researchers report today in Science Robotics. That’s big news not just for Curiosity’s science-gathering skills, but for the very idea of autonomy in space.
Here’s the problem with Mars: It’s too far away. It takes as long as 24 minutes to transmit a signal to or from the Curiosity rover—and that’s once a communication window opens up. So it’s not like operators can be sending messages back and forth to the rover all day. Instead, they upload a plan to Curiosity at the start of the day. Then, before it gets dark, the rover stops roving. But because of that delay, it isn’t as if Curiosity can send a picture of where it’s landed and wait for instructions on what rocks to sample. That means it’s wasting valuable science time.
So the AEGIS program helps identify those targets without Curiosity’s human handlers. What they’re after is bedrock—dirt is great and all, but bedrock is ideal because it’s still attached to the planet, stuck in the spot where it was formed. “What that means is that you know something about its context and the setting in which it was created,” says lead system engineer Raymond Francis of NASA, “and maybe even from its relationships to other materials, something about its history.”
And AEGIS loves it some bedrock. The rover begins by snapping an image and running it through computer vision algorithms, which look for contrast. “If you can find a sharp edge that you can close into a loop, you’ve probably found a distinct object,” says Francis. “And usually in these environments on Mars that means that you’ve found a rock or some coherent geological feature.”
Target acquired, the rover’s ChemCam fires a series of laser blasts at the rock and analyzes the composition of the vaporized material. And just like that, the rover has found and studied a Martian rock, all on its own. And it’s seriously accurate: It’s been able to successfully acquire the most desired material 93 percent of the time. And since its deployment, AEGIS has boosted ChemCam measurements by 40 percent.
Which is all the more impressive considering AEGIS is running on Curiosity’s less-than-superpowered computer. That’d be the RAD750 processor running at 133 MHz, using just 16 MB of RAM. “It’s the best computer we’ve got,” says Francis. “So we have to be kind of lightweight and lean and choose efficient algorithms.” (In fairness, the processor is also insanely tough, handling constant bombardment from radiation that would fry a normal chip, so be nice to it.)
Really, though, it’s the scientists on Earth who are the brains of Curiosity, no matter the fancy new algorithms and autonomy. “In general it’s important to remember that this is a science mission and so we’re not trying to replace the science team, we’re trying to give them better tools,” says Francis. “A smarter rover is a more useful tool to them, but it’s still the science team who’s in the driver seat.”
But lessons learned with AEGIS could well inform future space missions of all types. Full autonomy might not be good for a mission like Curiosity’s, but that won’t be the case elsewhere. Think of a mining rover that prospects for valuable minerals on its own. And really, the farther humans push into the solar system, the more indispensable autonomy will be. If you think the communication delay on Mars is bad, imagine trying to talk to a rover on Pluto—4.7 billion miles away to Mars’ 34 million miles.
So, autonomy: great for blasting rocks with lasers, not great for replacing NASA’s highly educated men and women in mission control. Solidarity, fellow humans.