In an experiment designed to show how robots can learn and evolve, researchers in Cambridge and Zurich programmed a robot arm – or “mother” – with an algorithm to create a device made out of blocks containing motors – its “child”.
The blocks are assembled into a structure by the robot arm and the motors are turned on. A camera detects how far the blocks are able to travel. The robot arm sees this, and then modifies the next “baby” to try and make it go further, learning from the mistakes and good traits of the last one.
This is all done without human intervention. The research was published in the journal PLOS One.