European engineers say they've developed a robot system that can "reproduce," building robotic "babies" that improve with every subsequent generation.
The researchers in Great Britain and Switzerland say the system can actually learn on its own, creating better robots each time by correcting its own mistakes, all without human intervention.
The "mother robot," as the engineers have chosen to characterize it, can test the performance of its children and make changes to the next generations to retain desirable performance traits while eliminating less successful ones.
The "babies" are simple plastic cubes containing electric motors, assembled by the "mother" robot arm in various configurations.
After assembly, the mother detects how far the baby cube assemblies can move on their own.
Then, without any input other than the initial instruction to build babies capable of movement, the mother reworks its designs to create succeeding generations capable of moving further and faster.
In tests, the engineers monitored the mother as it built 10 generations of babies.
Offspring in the 10th generation were capable of traveling twice as far as first-generation babies before running out of power, the researchers report in the journal PLOS ONE.
The robot is mimicking what happens with evolution in the natural world, they say.
"Natural selection is basically reproduction, assessment, reproduction, assessment and so on," says lead researcher Fumiya Iida of Cambridge University in England, who collaborated with researchers at ETH Zurich. "That's essentially what this robot is doing—we can actually watch the improvement and diversification of the species."
Using feedback gained from each generation, the mother robot was able to try new shapes and movement strategies for its "offspring," including some that a human designer would not have come up with, the researchers say.
Their chief aim, they explain, was not to create self-replicating robots but to investigate automated systems capable of learning, improving and adapting to new situations.
"One of the big questions in biology is how intelligence came about—we're using robotics to explore this mystery," says Iida. "We think of robots as performing repetitive tasks, and they're typically designed for mass production instead of mass customization, but we want to see robots that are capable of innovation and creativity."
Eventually, such "self-improving" robot systems could have applications in arenas such as manufacturing and agriculture, the researchers suggest.
"You can imagine cars being built in factories and the robot looking for defects in the car and fixing them by itself," researcher Andre Rosendo says. "And robots used in agriculture could try out slightly different ways of harvesting crops to see if they can improve yield."
You can view the robots in action in the video below.