Scientists produce robots that can artificially evolve and adapt to unfamiliar terrain with Darwinian thinking
Here’s some news that might send you scrambling to sign that petition to ban killer robots: scientists have for the first time produced robots that can evolve and adapt to new situations.
Using Darwinian selection, scientists at the Belgian University of Leuven found that they could artificially evolve robots, so they could self-organise the division of labour. In other words, different groups of robots automatically began specialising in different tasks.
The field of study is somewhat ominously known as “swarm robotics”, but researchers were inspired by nature, looking to replicate the complex societies developed by ants, termites or bees.
Comparing the robots’ ability to adapt to these social insects, the team wrote in journal PLOS Computational Biology that it showed for the first time that "self-organised task specialisation could be evolved entirely from scratch”.
Scientist Eliseo Ferrante commented in the study:
Such a division of labour could be achieved even if the robots were not told beforehand how the global task of retrieving items back to their base could best be divided into smaller subtasks.
The researchers say that robot swarms could be useful in future for space exploration, with their ability to adapt to unfamiliar terrain.
Whether this unfamiliar terrain includes, say, being sent 45 years into the past to kill Sarah Connor, the scientists unfortunately did not specify.