Cornell Creative Machines Lab
Nick Cheney at Cornell has developed a way for robots to evolve within days.
By Marshall Honorof
NEW YORK — Robots are simply more efficient than humans at certain tasks. They already excel at building cars, exploring distant planets and hunting for explosives, but it turns out that robots might also evolve much faster than their flesh-and-blood counterparts.
Nick Cheney, a Ph.D. student at Cornell University, presented his research at an Inside Cornell lecture on May 21. Cheney has developed a method by which complex computer simulations in a specific virtual environment — robots, by his definition — can evolve from selective pressures, just like animals in nature, but on a timescale of days instead of countless generations.
To demonstrate the technology, Cheney showed how a series of diverse but effective robots spontaneously evolved from a single, inefficient ancestor. He programmed a virtual environment with only one parameter: robots that moved faster would be able to produce more offspring. Therefore, the only selective pressure was speed (in the wild, Cheney compared this behavior to running from predators).
"Nature is amazing in how it designs things," Cheney said. "We want robots to interact with their environments as naturally as animals do." Cheney considers natural selection — the process by which biological organisms survive, reproduce and change over time to better suit their environments — to be a natural algorithm, extremely similar to what engineers use to optimize robots over time.
Cheney's initial robot, a shambling, cubic progenitor, was not much to look at. It barely stumbled along a straight line in no particular hurry. However, small mutations occurred in its offspring, and the fastest specimens bred with each other. As subsequent generations evolved, reproduced and died, the robots took on much more diverse appearances and began to speed across the screen.
The robots did not resemble traditional animals in any meaningful sense. Although they had come a long way since their initial boxy shape, they were still collections of small squares rather than sleek, curved specimens.
One robot resembled an accordion, constricting and stretching out as it made its way across the screen. Another, which looked like a wave about to crash, walked on three small points, almost falling over itself before finding its balance every few steps. Others walked on two distant legs, or maintained balance through rotating, top-mounted appendages. [See also: 10 Incredible 3-D Printed Products]
"These robots walk in ways we would never have thought of," Cheney said. Letting robots evolve without human oversight eliminates many of the preconceived ideas and biases that humans bring to the table by default. "We start from randomness, which is the way life started for us. Most of the random ones are pretty bad, but every once in a while you get lucky, and one will be better than the others."
Although these robots are confined to virtual space for the moment, that won't be the case forever. Thanks to the advent of 3-D printing, Cheney envisions a future where his robots could be powered by air, pressure-sensitive materials, electricity or even muscle, tissue and bone like real animals. Recent developments in 3-D printing have produced biologically viable heart cells, liver cells and even skull pieces.
"What we could explore with this is virtually limitless, which is what excites me most about it," Cheney said. Rapidly evolving robots with specific parameters could create everything from a better vacuum cleaner to complex search-and-rescue robots, but Cheney stresses that this is not the beginning of an adversarial relationship between humanity and its creation.
"In the future, we'll have more of a collaboration than a competition," he said. "Working together will be more fruitful than trying to take over the world."
- All-Natural 3-D Printers: Salt and Wood Can Be Used
- Top 7 Useful Robots You Can Buy Right Now
- 2013 Best 3-D Printer Reviews and Comparisons
Copyright 2013 TechNewsDaily, a TechMediaNetwork company. All rights reserved. This material may not be published, broadcast, rewritten or redistributed.