Descent of Robots

Researchers have observed the process of evolution by natural selection at work in robots, by constructing a ‘mother’ robot that can design, build and test its own ‘children’, and then use the results to improve the performance of the next generation, without relying on computer simulation or human intervention. Credit: University of Cambridge

At the University of Cambridge researchers are studying the evolution of species to copycat the process and "code" it into robots. 

They have developed a "mother" robot that can design and build robot children and observe their behaviour learning from it to improve the design of the next generation.

This is part of a new science, evolutionary robotics, where the objective is to create robots that can learn by themselves. This is quite a change from the idea of a robot as a machine that can do repetitive tasks without changing its behaviour. 

In the experiment the "mother" robot has been programmed to build its children out of several plastic cubes (up to 5) with a motor inside. Mommy builds 10 little robots at a time and then observe their behaviour that is basically dictated by 5 "genes" controlling the way the cubes are attached one another, the way the motors operate and so on.  It then uses the observation to determine who are the best performer and changes the genes of the less performing ones and produce a new generation of offsprings. 

The experiments have shown that there is a progress from one generation to the next and actually with no human (programmer) intervention at the end of the experiment the offsprings performed twice as good as the best performers in the first generation (they could move at twice their speed).

Interestingly, the announcement on the University of Cambridge website is titled "On the origin of (robot) species. Could it really be it?

Author - Roberto Saracco

© 2010-2019 EIT Digital IVZW. All rights reserved. Legal notice. Privacy Policy.

EIT Digital supported by the EIT