The beauty of evolution lies in two complementary forces: simplicity and complexity. From a simple rule—survival of the fittest—comes the astonishing array of critters that populate Earth. It doesn’t matter if you’ve got two legs or four legs or no legs at all, there’s no one right way to be on this planet.
Same goes, as it happens, for robotics. You and I are living on the verge of what you might call the Cambrian Explosion of robotics. Just in the last year, robots have been escaping en masse the factory and the lab to walk and roll and fly among us. Humanity has unleashed its own version of “life” on Earth, a sui generis genus that is evolving in ways that are fascinatingly similar to biological organisms.
Roboticists are honing their robots by essentially mimicking natural selection. Keep what works, throw out what doesn’t, to optimally adapt a robot to a particular job. “If we want to scrap something totally, we can do that,” says Nick Gravish, who studies the intersection of robotics and biology at UC San Diego. “Or we can take the best pieces from some design and put them in a new design and get rid of the things we don’t need.” Think of it, then, like intelligent design—that follows the principles of natural selection.
The caveat being, biology is rather more inflexible than what roboticists are doing. After all, you can give your biped robot two extra limbs and turn it into a quadruped fairly quickly, while animals change their features—cave-dwelling species might lose their eyes, for instance—over thousands of years. “Evolution is as much a trap as a means to advance,” says Gerald Loeb, CEO and co-founder of SynTouch, which is giving robots the power to feel. “Because you get locked into a lot of hardware that worked well in previous iterations and now can’t be changed because you’ve built your whole embryology on it.”
Evolution can still be rather explosive, though. 550 million years ago the Cambrian Explosion kicked off, giving birth to an incredible array of complex organisms. Before that, life was relatively squishier, relatively calmer. But then boom, predators a plenty, scrapping like hell to gain an edge. One theory for why is the development of vision, which allowed new species to better find mates and hunt food, setting off an evolutionary arms race.
The same may be happening in robotics, because robots are finally developing the senses that will let them understand the world and act on their own. Robots have for a while now navigated by building a map of their surroundings with lasers—the robocars of the 2004 Darpa Grand Challenge, which built the foundation for consumer self-driving cars, did this. But advances in computer vision are helping robots see better and better. For instance, a companion robot called Kuri navigates with lasers, but now also recognizes faces with machine vision. Getting robots to see well means training algorithms with lots of good data, and those algorithms will only get more powerful and efficient from here. It’s this literal and metaphorical vision that could set off a Cambrian Explosion of robotics.
And it’s not just that robots are adopting new senses—it’s that the proliferation of those new technologies is letting them settle into new niches. Earth features such an incredible diversity of critters in part because the planet has lots of different ecosystems for organisms to exploit (nature’s job market is generally strong), and the same thing is poised to happen for robotics. I mean, just think about the diversity of simple vehicles that already roll on four wheels. “There’s quite a range in terms of all the different kinds of sports cars and SUVs and offroad pickup trucks and tractor trailers, you name it,” says Jonathan Hurst, CTO of Agility Robotics, which makes the biped robot Cassie. “It’s going to be even broader I think for robots.”
Cassie walks on two legs, a nice generalist way of getting around. So the robot might work nicely as a base for a telepresence screen to walk around the office. “We’re going to want a robot that’s the same size to get around the factory floor and carry around 100-pound bags of something,” says Hurst. “There are going to be lots and lots of different niches, different sizes, different masses, different applications.”
The interesting bit about Cassie is that while it looks vaguely bird-like, Hurst and his team didn’t explicitly mimic anything in nature. What they were after was the most stable and efficient way to walk, and their math led them to a form factor that just so happens to look like a bird.
And we’re seeing robots adapting to every environmental niche available. A robot snake, for instance, might be good at exploring collapsed buildings, and indeed one went to work after Mexico’s massive earthquake in September. Wheeled robots are rolling on sidewalks delivering food. Other robots working alongside humans in factories don’t need to get around at all, but are…